How do you cultivate a wildly productive generative engine of design creativity. Especially one detached from your ego?
Design in this case means coming up with plans and specs for the thing you are making that fits the resource constraints. And the process of workshopping designs is where a designer coming up with multiple design solutions for the same problem and iterates upon them to better fit the constraints.
Steps in the workshopping process
Step 0: Play the game. Talk to folks in art, engineering, qa and production. Try to understand their concerns and issues.
Step 1: Write down your list of constraints and goals
Step 2: Come up with an idea that you believe in absolutely. That solves all the constraints. And delivers wonderful value. You’ve done it!
Step 3: Now put that intense passion aside! Socialize the design. Listen to misunderstanding. Come to terms with its flaws. What resonates? What do people start automatically building off?
Step 4: Do it all over again! Pick a new SEED, something out of all those chats that resonated with people. Build an entirely new solution around it.
Step N: Do this many more times! Until you’ve got a robust, doable idea that excites folks.
Some tips I’ve found that help do this well.
Be bold: Early on each new design can be wildly different. Cover a broad design space so you are working from a plentiful space.
Be safe: Cultivate a group of people you can share crazy ideas with in a safe fashion. If you don’t feel safe, you’ll struggle to be bold.
Be sure to fall in love: You need to sell this idea and provide design leadership around it. This could be the idea you ship! You should feel the Thrill when you describe it.
Take each idea two to three steps past obvious: First, think of the obvious thing. Now think of the next step beyond that.
Learn to fall out of love: Be comfortable putting your passion to the side. You can always make more beautiful ideas. You are infinite generative fire. No need to be precious. No need to get defensive.
When ideas start to converge, don’t settle for Frankenstein’s monster; a series of random parts stuck together by consensus. You want synthesis, not composition. A good idea’s elements supports the whole. A good idea generates more solutions, not more problems.
Make space: This process takes time. Give yourself the space to iterate.
Practicing the process
There are also things to practice. This is a skill you get better at over years.
Being faster: How quickly can you solve the constraints, fall in love, get feedback and do it all over again?
Being broader: Are your ideas crazy enough? Are you letting your brain be unhinged?
Listening: Does your listener’s face light up? What prompted them? Write that down! Can you riff off it? It could be a new seed.
Elegance: Solving more constraints with fewer pieces. Can you quickly come up with tight solutions to big problems? Learn to draw the goldfish with single stroke of your pen.
Converging: Knowing when to stop. Knowing when you’ve found the unifying thread.
Collaborating: The best ideas often involve a game of tennis where you bounce ideas off key collaborators. And you learn with each pass until something clicks.
Who created the final result? It doesn’t matter. Game design is a team sport and we are in this together.
A thing I’ve realized making non-combat focused games is how expensive combat mechanics often end up being in the end.
If you add combat…
“The basics” to get you in the door take immense effort. Enemy design, encounter design, player design, attack/defense systems, game feel, all the art and fx.
Then once you’ve got your combat system stood up, you are still competing again thousands of games that are roughly equivalent. It is the indie puzzle platformer swamp writ large. Lotsa substitute products. Trivial switching costs away from your game.
And of those 1000s, a handful of dominant titles are beloved hobbies. Players are fans of specific game feel often tuned over decades. The tiniest aesthetic deviation invalidates all your labor.
Balance your game for 12 months, still get comments like “feels floaty. sux.”
What experience do you really want to design towards?
And the amusing part is most of the clever differentiators that I care about exploring as a designer have nothing to do with combat. Story, progression, building, discovery, social systems, emergence?
You only really get to work on those after you overcome the Olympic-level challenge to ‘make combat perfect’. Often your inclusion of ‘a little combat’ ends up starving your actual design dreams of precious development resources.
So combat becomes this trap for many types of games.
You have to have it. Because that’s the expected core mechanic.
But it is going to be overly expensive.
And no one will like it.
And it will prevent you from making the systems you wanted to make in the first place.
I have a design tool I call “Don’t Solve the Hard Problem”
It consists of three steps
Identify: What is the expensive, difficult aspect of delivering on a highly desired user promise?
Clarify: What is the actual root of the promise?
Cheat: What is the the cheapest way possible to actually deliver on the root promise?
Example 1: Meaningful NPC interactions
Identify: Rich, Turing-testable AI is hard. Even in this era of deep learning, the results somehow feel empty. “Maybe in another 10 years…” <- The classic sign of a hard problem.
Clarify: But do all user actually want some weird chatty humanoid? When I talk to my players, what they really want is an emotional connection. A relationship. Someone to help and support and support them in return.
Cheat: In Cozy Grove, we didn’t solve NPC AI. We added an option to hug NPCs. They can refuse. But when they accept, it plays an animation. People feel an emotional connection. They feel support. That little hug is a meaningful addition to many of our player’s lives. It took less than a week to build.
Now, people working on the hard problem of NPC AI may be upset right now. The example I just gave is SO trivial. And isn’t the problem they are pouring their lives into solving.
But I didn’t have 10 years. So pivot. Solve an adjacent easier problem. It worked for our users.
(BTW, I’m constantly cheering on folks who ARE trying to solve the hard problems. It is the foundational work that moves the horizon of cheap solutions forward. Our ‘hug’ is built on digital animation tech that was a ‘hard problem’ of its own 40 year ago.)
Example 2: Massively multiplayer games
Identify: It is hard getting lots of players online at once. To this day I hear attempts at putting 1M concurrent players in the same space. Ugly bandwidth, CPU, replication, base costs and more result.
Clarify: What if player just want to be with friends? Sometimes lots of people, sometimes a few. What if “lots” and “few” and “friends” was a feeling that we could design for instead of a numerically large concurrency number?
Cheat: We’ve done experiments. Many times more than 30-50 people in a space feels like “lots”. Past 100, feel free to use bots or popular samples. No one can tell. Most successful multiplayer games deal primarily with smaller groups. What is the 4-12 player version of your game?
I remember old 1v100s discussions where 1M concurrent players was bandied about. Massive tech expenses resulted.
Yet most people played in 4-player matches. I’m always curious how the host wrapper could have been faked and that core 4-player feeling of being together enhanced.
Example 3: AR glasses
This one is more fanciful because the current discussion irritates me.
Identify: There’s a hard problem with a high FoV, high contrast, low latency, stable image that overlays the real world, works with the human eye and fits in glasses.
Clarify: The user fantasy I hear over and over again is “I want to remember the names of people I see”. There are others, but 8 out of 10 times this is the first use case real people can imagine for the tech.
Cheat: Why are we even focusing on the visual display for this problem? Seems totally secondary. One solve: You need a camera, a list of friends, face recognition, activation UI (voice? button?) and some headphones for feedback. All these things exist right now.
Ok, maybe the TRUE user fantasy is that this process is unobtrusive. Maybe we need to add a secret ring you can click and fiddle with to active the camera. Maybe we need the camera mounted on your head or eye tracking software to clarify intent.
Harder, but not the original hard problem
And what can you do with that boring platform? Translate text you look at. Identify objects. What the broad category of ‘finding out stuff about the world’ where the world is your affordance and looking is the pointer?
To me, this is still a hard problem. But at least it puts energy into something very solvable. And it isn’t some BS derived from fictional movies or tech executive wankery.
Imagine if this design thinking was applied to web3 or the metaverse. 😉
Look for the bodies
As a final note, when identifying hard problem, look for the bodies. Is there a history of smart people spending money and blood in this area with poor results?
Is there a slightly different problem to solve? One delivers what folks need, not what you’ve been told they want?
The joke goes: An expert game designer is 20x more effective than a newbie. They are correct 20% of the time instead of 1%.
Why are game designers wrong 80% of the time?
Sometimes they are wrong by a little. Sometimes by a lot. Is it poor planning? Are they morons? An expert painter does not produce a completely broken picture 80% of the time. Why is this so hard?
The feedback loop for creating other media is fast
I lay a lot of blame on the much larger gap between authoring a thing, experiencing the thing and revising.
Many types of media (like drawing or painting) allow for real-time ‘self-playtesting’ with the author as the playtester.
Game design does not.
When I draw, I am constantly engaged in a tight real-time iteration loop of authoring marks, viewing the marks, reacting to the experience as a viewer and adjusting the next steps. There are 1000s (often tens of 1000s) of feedback iterations.
Same goes for writing. There are larger editing passes that occur at lower frequencies, but even within those passes, I’m in a real-time create-experience-revise loop. The first draft is really the 5000th draft of the ‘self-playtesting’ process.
Now, when writing and drawing, I can’t predict exactly how someone-who-is-not-me will react. Death of the Author and all that. But an experienced artist and writer can often get within the ballpark for a familiar target audience. Sad scenes are sad. Happy pictures are happy.
The feedback loop for creating games is slow
Contrast that with games. 🙂 Some issues where the create-experience-revise loop breaks down.
Much longer iteration times. If I’m lucky it takes minutes to make localized changes and test them out. More typically it takes longer.
Due to interdependencies some changes can’t be fully experienced by the player until months later when all systems are fully in place. I just worked on a game where it took 1.5 years before we were able to test the basic flow and balance. This is common. Imagine having to paint a picture blind and wait a year before you can look at it and see if you painted it correctly.
Game developers often are corrupted playtesters. Many games involve mastery and knowledge. The designer, due to knowing what they know, becomes blind to issues new players will face. Empathy only goes so far, even when designers roleplay the ‘new player’.
Other systems (social systems, emergent complexity, proc gen, randomness, exponentials) are just hard to mentally visualize. We can plan them out, but the experience of playing them is often (deliberately) a surprise. There is no accurate ‘self-playtesting’ for these systems. A game designer’s has limited ability ‘play the game in their head’ and so real (slow) playtesting is required.
I don’t know of a perfect fix for any of this, but we have some tools.
Sketches: Movie makers (who also have extended pipelines) create low fidelity animatics that get to viewing the experience faster and cheaper. Game developers create prototypes that serve a similar role. It doesn’t work perfectly for all systems, but better is than nothing.
Genre expertise: Teams keep rebuilding games in the same genre over and over again. It might take years, but eventually you get to those 10,000 iterations. In large part, this is how expert designers even get up to that not-so-respectable 20% rate.
Community playtests: A large population of players + live development (early access, games-as-service) maximizes playtest feedback. Richer feedback can help counterbalance the slow iteration.
Content systems friendly to late-stage fixes: If you know that you are almost always going to be making big changes due to late feedback, you can build flexible pipelines that are easy to refactor. A proc gen system that creates 1000 levels constructed from modular components and centralized formulas is easier to tweak than 1000 handmade levels. Neither change is safe right before release, but at least the former is feasible.
Planning up front. There’s room for more waterfall style approaches. Particularly if you are reusing code, tools and have an experienced team. It works for things you know that you know. But this is surprisingly limited in other areas that comprise the bulk of game design.
So unlike writing or painting, the meta of game design is painstakingly building a process where you can iterate as quickly as possible, while making as few changes as possible, while still enabling big change to be feasible late in the process.
Recently I was designing the harvesting and crafting system for our Animal Crossing-like game Cozy Grove when I ran into a problem: picking up a stick is not that fun.
The core activities in a life sim are generally not full of mastery and depth. You chop trees. You dig holes. You pick up sticks. In isolation, each of these is dull. Our playtesters would harvest a leaf pile, get some sticks, and then put down the controller. They’d turn to me and ask “Uh, okay, where is the game?”
If we were following the standard advice on prototyping core mechanics, we might as well stop development right there. Clearly the core was not fun. We tried extending the loops out from 5-seconds (gathering), to 30 seconds (wandering), to 5-minutes (selling). No luck. My playtesting group hated the game.
Yet life-sims do exist! And they are delightful. Clearly there’s more to establishing value in a game than just perfecting a ‘fun’ core mechanic.
Discovering value chains
It wasn’t until we spent 12-months building out the rest of the game – the crafting, the decorating, the daily pacing structures – that players finally began to value picking up sticks. Because it turns out the value of sticks was entirely driven by their utility in reaching future goals. And if those future goals don’t exist, the sticks have no value.
You tend to see this scenario in high retention, progression focused games
The core mechanic is not the sole or even the primary driver of player value.
Value for a particular action comes from how it facilitates subsequent activities.
Often players engage in long chains of rote economic activity in order to reach their actual final goal.
Why you should care
Understanding how to generate meaning with sticks is not an idle concern! High retention, progression focused economic systems are at the heart of most games as a service (GaaS). There’s a huge demand for economy designers who know how to build and balance robust game economies that provide rich value to players.
Getting your economy design wrong costs time and money. Very often, it can kill the game. Yet game economies are also a rarely discussed black art. So it is hard to know where to start. And hard to hold constructive conversations with your team. There’s an inherent complexity to the topic that makes matters even worse.
So let’s try to improve the situation.
What this essay covers
We’ll cover the following.
Chapter 1: What is a value chain?
Chapter 2: Balancing value chains
Chapter 3: Architecture of multiple value chains
Chapter 4: Establishing endogenous meaning in games
The whole essay is around 30+ pages and can be a bit technical. Feel free to take it slowly. But if you are interested in game economy design, this is a good crash course.
CHAPTER 1 – WHAT IS A VALUE CHAIN?
We’ll model game economies and associated activities as endogenous (self-contained) value networks. These networks are composed of value chains. The value chains combine to form a full faucet-and-drain economy.
Basic structure of a value chain
This is the shorthand I use when jotting these out on paper.
You get a stick!
Which lets you make a lamp
Which lets you decorate your house
Which satisfies your need for self expression, the ultimate motivational anchor for wanting the stick in the first place.
Notice the structure
Each node contains an output of some (currently unspecified) action.
The nodes are connected to one another in a linear fashion that’s easy to read. No strange loops or spaghetti-like diagrams.
The chain terminates with an anchor node representing player motivations.
There’s a lot that’s not specified here in the shorthand version. We’ll get into more verbose versions below.
But you can see some useful traits
By jotting this down, you are forced to consider the direct purpose of each resource.
And how it relates to ultimate player motivations.
If the chain is broken, imbalanced or obfuscated in some way, players will stop finding value in the early steps of the chain.
Inputs and outputs
Now let’s look at a more verbose description of a value chain. In practice each node is composed of three elements:
Action: What the player (or the game) is doing to cause a change in the world.
Inputs: Resources that the action requires. These can be tangible resources or abstract concepts like time.
Outputs: Resources that are the result of the action. Again, they can be concrete or abstract in nature. The anchor node is always abstract since it represents an internal psychological state.
The diagram above is pretty, but hard to quickly put together. In practice, I use a text-based format that can be typed out in any basic text editor. Feel free to adapt the formatting to your project; it is the ideas that matter.
In purely text format, we get something like:
ChopTree (-treeHealth & -player time) -> +stick
Craft ( recipe & -stick & -rag ) -> +lamp
Decorate ( lamp, other decorations ) -> Decorated Space
The , symbol represents options that are valid inputs. Decoration can occur with the torch OR any other decoration.
There’s no concrete new resource here that is produced but we do get a decorated space as an output.
Step 4: “Decorated Space -> Self-expression anchor”
Finally, the player serves their goal, which is to express themselves. Self-expression is a strong intrinsic motivation for some players and acts as the anchor for the entire chain.
Key concepts when working with value chains
Now that you have a definition of value chains, let’s look at how they tie into some other important game design concepts.
Value chains are one way of structuring your game’s internal economy
Internal economies refers to the practice of modeling economies as a network of the following basic operations
Tokens: Resource tokens that flow between various nodes in the network of player or system operations.
Sources: A node that creates new tokens and add them to the flow.
Pools: Nodes that accumulate or hold some number of tokens
Transforms: Nodes that transform of tokens into other tokens
Sink: Nodes that destroy those tokens.
For a rich description of how internal economies work, read Joris Doman’s book Game Mechanics: Advanced Game Design. He goes into common design patterns and explains the ideas in more detail.
You can describe almost any economic system in a game using these basic elements. But that flexibility also can be overwhelming and hard to communicate.
Value chains are a specific sub-case of an internal economy. They make use of all the basic operations but in a far more restrictive manner that makes both the construction of economies and more importantly, their subsequent analysis easier.
Value chains are a form of “Faucet-and-Drain” economy
In a typical real-world economy, tokens circulate in enormous pools that slosh back and forth due to feedback loops and emergent market dynamics. These are enormously complicated and difficult to visualize.
However, value chains focus on a simplified cartoon economy known as a faucet-and-drain economy, defined by strong sources and strong sinks, limited object lifespans and limited circulation of goods.
A faucet and drain economy (like the one visualized for Ultima) might seem complicated at first glance, but it has some greatly simplified attributes
Faucets: Resources in the game are generated from nothing as needed. They are virtual so we can make as many as we want (or as few)
Transforms: The resources flow mostly one way into a series of transforms to produce various other elements players desire. These are all designed game activities.
Drains: We get rid of excess materials. Again, they are digital so there’s no unexpected externalities like pollution or landfills. We press a button and boop, they are erased from existence.
These faucets and drains map directly onto the various portions of the value chain.
Early stage of the value chain has sources: Players perform actions (the core gameplay!) and generate a steady flow of base resources.
Mid stage of the value chain transforms resources: Those resources are transformed into a very small number of intermediate resources.
End stage of the value chain has sinks: Finally players pay resources into sinks that help them gain access to whatever their anchor motivations might be. As a result, there is a steady stream of goods being destroyed.
Faucet-and-drain economies have some really useful attributes for economy designers.
Constant demand: There’s a nice velocity of goods flowing through the economy and out via the sinks. This means we can easily incentivize players to continue engaging in gameplay actions that generate resources.
Limited feedback loops: There’s limited pooling of excess resources. This leads to simpler dynamics and fewer unexpected feedback loops.
Easier explanation: Real economies are complex and hard to talk about. A faucet-and-drain economy is much easier to explain to players. This lets them form models of cause-and-effect and helps with their long term planning and engagement.
Easier balancing: You can usually trivially balance sources and sinks against one another. If there isn’t enough of a resource being created, the designer can tweak some number so player actions generate more. Or decrease the cost of sinks. If there is too much of a resource being generated, the designer can increase the cost sinks or reduce the sources.
Game economies as cartoon economies
Note: You don’t see many pure faucet-and-drain economies in the real world. All real-world economists need to deal with the messy reality that real-world extraction of resources and making of objects is incredibly expensive. You can’t wave a magic wand to increase some source of scarce goods. So instead, we see more circular economies where limited rival goods are created infrequently and then circulate within a competitive market for a long period of time. Your modeling must include supply chains, warehousing, environmental externalities, transaction costs and more.
Games economies are special since there is zero cost to creating, transforming and destroying new digital resources. If we want every person in the game to get a puppy, we snap our digital fingers and it is done.
This means every economic feature associated with supply chains, transaction costs and various externalities is a design choice; we include them if they make the game better. As digital economy designers, we can use these special powers to make our job as game economy designers easier and the experience of playing the game better for our players.
Value chains are structures for creating demand
A node creates demand and value for players to engage with earlier nodes in the chain. You can say that a node ‘pulls’ resources up from those early nodes.
The powerful sinks at the end of each value chain acts to maximize flow of resources through the chain. Because we usually want clear systems of cause and effect so players can easily plan ahead, building murky pools of inscrutable assets can hurt gameplay if not done with care.
Big pools reduce pull. However, see “The emotions of scarcity and abundance” below for more detail on how planned scarcity and abundance (within a narrow band of outcomes) can drive desired player emotions.
Value chains connect to real world motivations through Anchors
Traditionally, when we think of value, we often think of valuable goods as those that serve a physical need like food or shelter. However, in games, there is no actual need for food or shelter to satisfy. A game will not feed you if you are hungry. Instead valuable digital goods are those that serve a player’s psychological needs. Game food in a game like Valhiem fulfills a player fantasy of survival (mastery over the environment), not actual hunger.
When designing value chains, anchors are how we define and represent these psychological needs. And by inserting them at the end of each value chain, we call out how meaning is carried through the earlier stage economic nodes. A powerful anchor can be a big reason why earlier nodes have value. If you don’t identify how each node serves your psychological anchors, you’ll very likely end up with disconnected nodes of isolated atomic meaning.
Other sources of meaning: Each node can always generate meaning. It can contain a beautiful interaction. Or a delightful puzzle. Or moment of insight. By no means am I saying that only economic value networks provide meaning or value! However, if you can connect these small moments of value together by a self-reinforcing value network, you build a self-consistent space for players to form and pursue meaningful goals.
Possible anchors: There are lots of motivations that act as anchors. For example SDT (Self Determination Theory) lists
Autonomy: Do you accept the decisions you have made?
Competence: Do you feel like you are gaining mastery or competence in your actions
Relatedness: Are you supported and do you support others in a pursuit of self determination?
Or you could reference work by Quantic Foundry, who has tried to map out some of the player motivations from popular games.
Player interviews: Ultimately, these models are only a starting place for finding your game’s anchors. You’ll discover that needs are nuanced and perhaps best uncovered by talking to your players and finding what resonates with them. When you hear that your game changed someone’s life for the better, don’t roll your eyes. You are likely observing an anchor for a value chain.
I remember once someone told me that my cooperative factory building game helped restore his faith that people can be good. And how he made lifelong friends just through playing the game. Those shared goals, trust and long-term friendship are powerful anchors that made the atomic actions of placing road tiles and cranes very meaningful.
The deep meaning behind value chains is this: By understanding anchors you start to understand how your game provides tangible value to your player. Games are never ‘just games.’ People play them (and keep playing them) because games add real value to their lives. You need to design with that goal in mind.
Can economic systems serve intrinsic motivation?
At the most basic level, extrinsic motivations are when you feel forced to do an action in order to attain some other purpose. Intrinsic motivations are when you want to do an action for the sake of the action itself.
It is around this point that folks trained in the pop-psychology pits of YouTube game design start worrying that economic systems of this sort promote coercive extrinsic motivators. The actual answer is complicated.
Some things to keep in mind that are usually not discussed in a typical like-share-subscribe rant.
Extrinsic vs Intrinsic motivation is a spectrum: It is not a binary. A lot of things we do are a little extrinsically motivated and a little intrinsically motivated. So if you are desperate for black and white morality, be aware this is a topic that is mostly shades of gray in practice.
Individual perception matters: People often move from being extrinsically motivated to being intrinsically motivated for the same exact action and reward as they incorporate the action into their personal feelings of self-determination. For example, at one point I made a french press coffee each morning to wake up and get my caffeine. It was extrinsically motivated, rote behavior. But then I started thinking of myself as a coffee drinker and over time the rote activity turned into a ritual that I truly enjoy. People shift and change. Perception is often more important than the exact mechanics or numbers involved.
Intrinsic motivation is often a journey: The story about coffee suggests another truth. Players don’t start out intrinsically motivated. Often they are just playing around, following once bright sparkly to the next bright sparkly. Overtime they discover how a set of rote actions serves one of their unmet deeper human needs. This can take time and learning! But by the end of their personal journey, the previously rote actions are transformed. They become time spent with purpose and intention.
Why you choose to perform a rote action as well as your level of personal buy-in to this choice have a huge impact on whether or not an activity is seen as intrinsically or extrinsically motivated. If you are interested in this topic, I highly recommend doing a deep dive into SDT since it is one of the few experimentally verified models of key factors involved in shaping motivation.
Design each action node to support feelings of autonomy, competence and relatedness. The more each moment in the game supports feelings of self-determination, the more likely players feel intrinsically motivated.
Ensure each anchor supports feelings of autonomy, competence and relatedness. If your ultimate anchors are tied to materialistic numbers going up like “Make as much money as possible”, you really aren’t supporting any of the key factors related to intrinsic motivation.
“Mastery” is often a short term motivator: Historically games have focused on helping players feel competence by teaching them novel skills. For example, you learn to double jump for the first time or beat a new boss puzzle. But eventually players learn the skill and chunk it into a rotely executed tool. True evergreen mastery mechanics are rare and expensive to invent. But ‘infinite mastery’ doesn’t need to be the goal of every mechanic in your game! Because it turns out that rote competence in service of other unmet needs still triggers feelings of competence! Think of mastery in terms of the player creating cognitive tools they can then apply to serving higher order needs.
Things to listen for in playtests
Coercion: Do players say they feel coerced into doing something? If you hear this, you are leaning too heavily towards extrinsic motivators.
Changes in perception: Do players say “This didn’t meet my initial expectations, but now I really enjoy it.” That suggests players are transitioning from extrinsic motivation to intrinsic motivation. Ask them why they enjoy it. There’s a very good chance you’ll discover some powerful anchors in your design (that you can then amplify or reveal sooner!)
The big picture: There’s a fun study that suggests games that lean heavily on intrinsic motivation tend to improve player’s well-being. And they tend to have longer term retention and improved monetization. While games that lean heavily on extrinsic motivations tend to harm player’s well-being.
So even though the topic is messier than suggested by internet moralizing, it is still worth building in the factors of intrinsic motivation into every step of each value chain.
CHAPTER 2 – BALANCING VALUE CHAINS
Why you balance your economy
One of the critical jobs you’ll perform as an economy designer is making sure your economy is balanced. As game developers, we are selling amazing experiences and a poorly balanced economy leads to a crappy player experience.
From the player perspective, an imbalanced economy produces complaints that look like the following:
X activity or resource seems pointless.
X is boring
I didn’t even notice X
I don’t understand why I’m playing.
These mundane phrases are some of the most important pieces of player feedback a designer can hear. Your players are not being stupid. They are giving you incredibly valuable signals on what is wrong with your game.
Role of value chains in economy balance
There are many potential root issues that drive this sort of feedback. The trick is finding them. For example, an abundance in one location might be driven by a lack of sinks further down the chain. If you don’t know the structural dependencies of your economy, you’ll struggle to pinpoint the root cause of a player’s report.
Value chains provide an analytical framework that helps you do the following:
Define each resource in the game and why each is valuable.
Define structures for how resources are relate to one another in a meaningful way.
Analyze, pinpoint and fix issues where resources or activities are not valuable.
Every prototype you make starts out poorly balanced. And then you iterate on the balance to make it better. Value chains speed up iteration by simplifying the underlying problems and helping you identify and classify observed problems faster. They help you reduce the cost of fixes by targeting specific problems while limiting ripple effects.
Balance from the technical perspective
From a technical perspective, we can define a balanced value chain as one where there is a strong enough set of anchors and associated sinks to consistently pull resources up from all nodes along the chain.
The player is motivated at each node to perform game activities in order to reach subsequent nodes (and ultimately the anchor)
The player doesn’t face an overabundance of a particular resource that swamps sink or makes exercising an earlier node’s activity meaningless.
The player doesn’t face extreme scarcity of a much needed resource that makes grinding an earlier node laborious or irritating. This can lead to pacing delays or grinding burnout.
There are a few key steps to balancing a value chain. Each of these is a major topic we’ll cover in detail.
Step 1: The structure of the value chain must have clear links of cause and effect carrying over from each node all the way to the anchor.
Step 2: You need to identify the types of sources and sinks used in your value chain.
Step 3: You need to match the power of your sinks with the power of your sources. For example exponentially increasing sources should be matched by exponentially increasing costs on sinks. Mismatches here result in nearly impossible to balance economies.
Step 1: Debug the structure of the value chain
At the most fundamental level, a value chain is a connected series of economic actions. If the links in the chain don’t connect to one another at a structural level, the chain fails. This is super useful!
We can look at the structure of any specific chain and quickly identify structural errors without taking into account the massive complexity of the whole economy.
We can often further focus on a single node in a single chain to identify the issue with great specificity.
Errors at the structure level are great to catch early since they often result in economies that are impossible to balance. Defining and debugging the structure of your chains is a wonderful pass on any economy design task.
Issue: Break in the value chain
The most common root issue is that there is no subsequent link in the value chain! Most actions in games quickly get mastered, chunked up and turned into a rote task. If there isn’t a reason to do the action, it becomes as meaningless as a disconnected doorbell.
Write out the value chain for this action.
Make sure there’s a consistent chain of nodes all the way to an anchor.
Issue: Lack of sufficiently compelling anchor
Identifying compelling anchors is a rarer skill. So many designers just leave out this step entirely. However, the game then falls flat and they don’t know why.
Your intrinsically rewarding anchors are often very related to your game’s core pillars or promises.
Do the exercise of asking what players really want out of your game in terms of need fulfillment or core motivations.
See if you can have your value chains directly contribute to these ideas. Your game will become stronger. You’ll also gain a culling device for eliminating features that don’t serve the pillars of your game.
Issue: Visibility on the chain of cause and effect necessary to reach the anchor
In games, players engage in interaction loops that teach skills on how to manipulate the world. Interaction loops and arcs (also known as skill atoms or learning loops) are the fundamental iterative sequence of modeling, deciding, acting, processing and responding that occurs within any computer mediated interaction. It is the heart of any interaction design.
Here’s how interaction loops map onto value chains
Each interaction loop directly corresponds to the action element inside node of the value chain. For example, there is an interaction loop about learning to harvest leaves. And that maps onto the value chain node about harvest leaves and getting sticks.
Exercising an interaction loop yields emotional reactions. There’s evocative stimuli (ooh, pretty jewels go pop!), mastery, autonomy and more.
A player exercises an interaction and learns cause and effect. A chunked skill always results in a lesson, or cognitive tool for how an interaction can manipulate the game. For example, players learn that if they harvest a leaf pile, they’ll get sticks.
The player can then use their new acquired tool to pursue goals. This corresponds to a subsequent node in the value chain. For example, if a player wants to build a decoration, they now know they need a stick.
Interaction loops are recursive in nature and occur at pretty much every level of gameplay. I’ve written a lot about them over the years and they are essential to almost every part of my design practice. But that’s far too much to cover in this relatively small essay that I’m desperately trying to keep focused on game economies. If you want more information on interaction loops, check out this presentation: Interaction Loops – Public
The important part is there are many ways a specific interaction loop can go wrong. The interaction loop may have the wrong affordances. The game could provide poor feedback to the player. The player might not have learned foundational skills. Etc, etc.
From the perspective of analyzing value chains, you should know that a failure inside a single node of the chain can destroy the value of all subsequent nodes along the chain. Knowing these dependencies can help you backtrack and find the root causes. If you can track a big economic issue down to a single interaction inside a single node, you can make far more targeted changes.
Issue: Visibility of the anchor
You may have a strong anchor for a value chain, but only long term players end up figuring it out. And new players, because they don’t see how the game fulfills their needs, decide to leave early.
Identify your value anchors and tell them to the player at the very start of the game. This is your player promise.
They won’t be able to experience the satisfaction of the anchor immediately, but they’ll know what they are working towards. And this should give them a long term goals and perspective on the long term payoff of current tasks.
The player promise can be couched using a narrative frame. For example, a need for completion and accomplishment is often couched as ‘beat the game’. A need for dominance and mastery is often couched as ‘beat the final boss’. These simple frames help contextualize the abstract psychology of an anchor as a familiar concept. This can provide enough visibility on an anchor to justify earlier actions.
Issue: Weak player motivation associated with the anchor
Motivation and their associated narrative frames are not universal! Many players don’t care about dominance or mastery. In our life-sim, Cozy Grove, players were actively repulsed when mastery elements were experimentally added. If you present the wrong audience a game about beating the final boss, they will leave immediately because their true needs are not being met.
Talk to target players. Tell them your player promises. See if they are excited! If you are a new developer pound into your head that there is no universal gamer profile. Nor is there a game that is perfect for everyone. (You’ve been lied to if you believe this)
If they aren’t excited, you need to either find a different set of player promises or a different target audience.
Don’t be afraid to workshop your player promises until you find a strong audience fit. A mismatch here can cause your game to fail before you even start.
Step 2: Identify types of sources and sinks
In more freeform descriptions of internal economies, there are innumerable ways of adding resources and extracting resources from the economy. Once you start including feedback loops, pools, conditionals on actions and other attributes of an internal economy, you might as well have written a full Turing complete simulation. Such a system is difficult to explain, difficult to reason about, and difficult to balance. (Check out Machinations.io if you are interested in exploring what these simulations can look like. It is a wonderful tool.)
In my personal practice building game economies, I’ve hit upon a relatively robust simplification where I categorize source and sinks into a few common categories. There are certainly edge cases that these types will not cover. However, by restricting your economy design to well-defined and easily manipulated components, you make balancing far easier.
In this section we’ll talk about how to approximate complex economic structures with various curves.
The 5 major types of sources: Capped, Trickle, Grind, Investment and Random
The 4 major types of sinks: Fixed, Repeatable, Exponential, Competitive
The 3 big balancing challenges: Scarcity, Abundance, Variability
Source Type: Capped (Constant)
In this type of source, there’s a fixed amount of the resource that comes into the game via some action node (or nodes).
Player completes an action. I sometimes use the metaphor of ‘turning a crank’ where the player needs to execute a full interaction loop of: mental model -> decision -> action -> rules processing -> feedback -> updated mental model and resources.
The total amount that comes from the action is fixed.
Executing the action again (if even possible) does not provide more of the resource.
Some variations on capped sources include:
Capped sources are one of the most common sources, especially in single player games with a fixed completion point. They tend to be used in easy to control systems, but can be a little brittle. We’ll get into that more when we discuss numerical balancing.
Source Type: Trickle (Linear)
In this type of source, there’s a fixed rate of a given resource coming into the world during a given time period.
Again, like with capped sources, a trickle resource delivers resources whenever an action node is completed.
However, unlike capped sources, you tend to get a consistent amount every single time the action is repeated. Forever.
Accumulation challenge with trickle sources: On larger time scales, you can accumulate an infinite amount of that resource. Say you earn 10 gold every day for signing into the game. After 10 days, you have 100 gold. After 100 days, you have 1000 gold. After 2 years, you have 7300 gold.
When players have an excess of a resource, they are less economically motivated to engage with the production nodes of its value chain. Though they may still perform related actions for the intrinsic joy of it, the marginal value of gaining an additional resource is low.
We’ll keep seeing excess accumulations show up as one of the failings of an imbalanced economy.
Variation: Limited actions per time period: In Animal Crossing, there are 6 rocks that spawn in the world and can be mined once a day. This is still a rate limited source, but the limit is placed on the actions the player can perform, not the amount of resources produced.
Variation: Capped pools fed by a trickle source: A common type of trickle resource is energy in a F2P mobile title. Energy recharges each day and can then be spent on a limited number of actions. Some elements of this patterns
The action of the energy production node is simply ‘waiting’. Time passes and you automatically get more energy. (You can pay, but that turns this into an investment source, which we’ll cover in detail below)
There’s a capped pool, which is a pool that holds the energy resource. It is capped in that it only holds some maximum amount. After reaching the cap, any additional energy is lost.
Capped pools are one partial solution to the accumulation challenge. In our gold example above, imagine that gold feeds into a treasure chest that can only handle 100 gold. If you wait 10 days, you’ll have 100 gold. If you wait two years, you’ll still have 100 gold.
Capped pools are unfortunately not a complete solution. Someone who diligently empties the pool every day still will be able to spend all 7300 gold over two years. So you still need a mechanism for dealing with excess.
Source Type: Grind (Linear)
A grind source is one where players can spend near unlimited external resources such as time or money to increase the amount of a given resource. Again, you’ve got an action the player performs on a node that generates resources. But they can grind that action by performing it as many times as they desire.
Though on the surface this looks a lot like a trickle or capped source, from a balancing perspective it is very different.
The total amount is mostly uncapped. It is limited only by how much a player wants to grind overall.
The rate is also mostly uncapped. It is limited by how much a player wants to grind in a day.
The most common example is that the player invests more time by repetitively performing an action again and again. Though time is limited in reality, in practice we often balance our games by assuming a certain moderate level of engagement. And someone who plays 14 hours a day, 7 days a week can grind out surprising amounts of a resource.
Variability Challenge: The big problem with this source is that it is highly variable, which makes it hard to balance. A player could not grind at all. Or they could grind 18 hours a day for 300 days. In one case, you’ve got scarcity. In another case, you’ve got overabundance. Both players will complain that the economy is poorly balanced.
The pattern of play varies how much a grind source produces. In the chart above, we see a bump for player 1 during the weekend. They may experience a huge glut of resources as a result.
It is often good to convert this into a capped or trickle source. Or pair it with an exponential sink.
Source type: Investment (Exponential)
A common structure in internal economies is the positive feedback loop.
Player does an action
This gives them resources.
But these resources ‘feed back’ into the original action. They can invest the resources to do more of the action.
Which in turn gives them even more sources that lets them to the action even more.
Positive feedback loops result in hard to balance economies.
Early on in the investment cycle, these sources produce small quantities of a resource.
Later on, positive feedback produces exponential growth of a resource.
But this exponential growth oly happens if the feedback loop is being actively exploited by a smart player.
So we end up with scarcity early on and then are hit with abundance and variability later. It can be a huge pain.
An example of an investment resource might be fruit trees in Animal Crossing. When you start out, you feel great harvesting a single fruit tree.
Simplification: Treat Positive Feedback Loops as Exponential Sources: For years, I’ve been relying on a straightforward simplification: I treat positive feedback loops as exponential sources. I design defensively and assume the worst case scenario where feedback loops are going to get out of control for some players.
This cartoon model of a positive feedback loop has several benefits
Instead of dealing with complex, tricky to communicate diagrams that chart out the exact structure of a feedback loop, we can just say a particular node is an investment source. This lets us continue to deal with the economy using targeted value chains.
When we get to balancing sources and sinks, it unlocks clear numerical tactics for sopping up abundant resources. Instead of some mysterious dynamic system of emergence, the source becomes just another common type of math curve.
Variation – Increasing the starting baseline: We can help eliminate scarcity during early stages of an investment source by adding a trickle source to fall back on.
Animal Crossing’s fruit is not a pure investment source. Instead, they start you out with a set of ‘wild trees’ that let you harvest at least some baseline quantity of fruit each day even if you haven’t engaged with the investment loops.
Source Type: Random (Noise function)
Some sources come with high variability. The most common of these in video games is a loot drop table, but almost any game that uses dice to determine resource rewards is using a random source.
Random sources have a distribution of outputs: They can be a normal curve, exponential distribution, pure random noise or some other histogram. These will usually have some sort of central tendency where on average you can have a typical result.
Random sources are really just a noise function applied on top of one of the other source types. So you can have capped, trickle, grind or investment sources with randomness.
Simplification: Use the mean of normal distributions and convert to a less random source type: I tend to work with average outcomes when looking at how random sources contribute to the economy long term. This turns a random source into one of the other sources (capped, trickle, grind, investment).
This simplification doesn’t work for very short term variability balancing, but can be highly effective for understanding scarcity and abundance in longer lasting games.
Variation: Constrained randomness: You almost never want pure randomness in an economy. In your million of players, there will be that one person who rolls 1s for most of their game and attains 1% of the progress of the typical player. The system isn’t broken. Sometimes true randomness results in crap outcomes.
If possible, use systems like ‘drawing-without-replacement’ (decks of cards with a discard pile work this way) or various pity systems that guarantee a drop after X draws. These ensure that the outlying experiences aren’t substantially different from the average experience.
Again, we aren’t interested in techniques that allow you to balance any system. We are interested in building systems that are easy to balance.
Sink Type: Fixed (Constant)
Now we get into sinks. These extract resources from the game and thus limit accumulation and pooling. You’ll see that these fit into categories very similar to sources (constant, linear, exponential)
The simplest sink is the fixed sink. When an action on a node in the value chain occurs, a fixed amount of resource is removed from the game. This is not repeatable. This is the mirror of a capped source.
There are lots of examples of fixed sinks
A powerup you can purchase once. This takes a fixed amount of currency out of the economy.
The one time cost in XP to earn the next level in an RPG. This takes a fixed amount of XP out of the economy.
A boss you can beat a single time and in the process it uses up healing or mana potions.
Most fixed-length games make heavy use of fixed sinks. You put them all in a spreadsheet and tally them up. This tells you how much you can give out from capped sources.
Sink Type: Repeatable (Linear)
The mirror of a trickle source is a repeatable sink. Every time the action for a node is performed, a fixed amount is removed. However, unlike a fixed sink, the action can be repeated multiple times.
Some common examples of repeatable sinks
Damage being done to someone’s health bar. Each time the attack repeats, the same amount of health is lost.
The crafting cost for a crafting recipe that can be crafted multiple times.
A lamp in Valheim you need to regularly refuel or else it goes out.
The cost to buy a consumable item in the store that replenishes each day.
The cost to purchase a tree in animal crossing.
Why not distinguish trickle sinks and grind sinks? You’ll notice that there’s no mirror sink to the grind source. You can absolutely have a trickle sink that only allows a certain amount of some resource to be destroyed in a given time period. Or a grind sink where players must grind to remove more of a resource.
In practice however, these distinctions tend not to matter too much. Repeatable sinks are naturally limited by the supply of a resource. So we don’t get the specific runaway cases like ‘grinding’ that need special attention like we do with sources.
Sink Type: Exponential (Exponential)
The mirror of the investment source is the exponential sink. In this sink, to get the next incremental (linear) increase in output, we logarithmically or geometrically increase the input quantity. This means there’s always room to sop up more.
Some examples of exponential sinks
Each additional level for an RPG character costs exponentially more than the last level.
In an idle game, each upgrade to an idle resource generator costs exponentially more than the previous upgrade.
Sink Type: Competitive (Adaptive)
There’s a specific type of sink that doesn’t have a clear mirror source. A competitive sink is a form of adaptive sink. In a competition between multiple players, whoever puts in the largest amount of a resource gets the largest prize.
Pro: The nice thing about this sort of sink is there’s no top end so it can suck up lots of resources.
Con: However, it can only be paired with competitive motivational anchors. And only a tiny percentage of the population is motivated by competition (mostly young males). So there are limited types of games you can use this.
Examples of competitive sinks
Guild vs Guild competition in a game like Clash of Clans
Armies battling in an RTS game.
There are lots of variations of this type of sink
Races: Players try to reach a specific goal. Whoever reaches it first, wins. There can be vast and expensive chains around training and other improvements to enhance your ability to get ahead of others.
Leaderboards: There are more than two players competing and the positions are ranked relative to one another. So someone comes in 1st place, 2nd place, 3rd place, etc. Often the rank is measured within a league or session window.
Mixing and matching sinks
All these types can also be mixed and matched. Idle games use leveling as a repeatable sink whose cost increments exponentially each time you level. Leveling can be fixed at a fixed number of levels. So short term, a sink is exponential, but long term it is fixed.
Like sources, it helps to classify a sink within a given time window.
Step 3: Match power of sources and sink
You can essentially classify these sources and sinks according to their power
Constant: These are your capped sources and fixed sinks. This is the lowest power. (x^0)
Exponential: Investment sources, Exponential sinks. This is the highest power. (x^1+)
Adaptive: Competitive sinks (A>B). These are special cases.
When balancing a value chain, it is immensely helpful to have lower power sources feed into equal or higher power sinks.
The challenge of overabundance
A slight amount of periodic scarcity or feelings of abundance can provide necessarily emotional variation to your experience. However, if you don’t balance your sink types, then as time goes on, a resource starts to accumulate in a pool somewhere and can’t be spent. This creates overabundance.
When there is no chance of scarcity or when there’s no use for those pooled resources, people stop caring.
It may be trivial to satisfy or even exhaust their motivational anchors. If you are motivated by status and you can instantly buy all the high status clothing, why bother continuing to exercise that value chain?
There is no longer a pull on the earlier nodes in your value chain that produce that resource. If you have all the sticks you’ll ever need, why bother ever harvesting another dirt pile?
In CRPGs where you can sell things to vendors, you end up with millions of gold. But you have max level equipment. So what is the point?
In MMOs, players eventually have access to a 1000 useless +10 swords. This is known as mudflation and creeps up on games over the course of years. Why keep killing rats?
Of course not all value chains need to last forever.
If you’ve created a fixed sink for a given value chain and thus have planned its eventual obsolescence, then it may be fine to let resources accumulate.
Or the player may enjoy messing about with a large amount of resources in a creative mode. They find this intrinsically rewarding and don’t need the careful scaffolding of motivations that a taut value chain provides.
But intrinsic motivation is something that most people need to slowly work their way towards. They benefit from practicing an activity for long enough to know they enjoy it and eventually understanding how it serves their needs. This onboarding via explicit affordances and feedback seems particularly important when a player is operating within the artificial cartoon value structure of a video game.
So reducing overabundance, at least in the early portion of a game and especially in games without creative anchors, is usually a good starting point for balancing your economy.
The good news here is that big desirable sinks make it trivial to balance most faucet and drain economies. They create a need for resources which in turn causes the players to engage in actions all the way down the value chain.
Balancing a fixed-length game
The easiest example is the fixed-length game. These are ones where you can map out the economy for the entire play experience, from start to some finite completion. These are good learning projects for new economy designers.
List your sources in a spreadsheet for each node of the chain: If you have a fixed-length game composed of a series of capped sources, make a spreadsheet and add up those resources the players will encounter over the course of the game. You can also sum up trickle sources since you know when the game will end.
List your sink for each node of the chain: Now make sure you have a set of fixed sinks that consume those resources. Include repeated sinks as well. Again, add them up!
Golden path modeling: For trickle source and repeatable sinks you’ll need to decide how many times an ideal player interacts with each. This ‘golden path’ won’t be followed by every player, but it helps you approximate how much they’ll earn and spend.
When picking which sources and sinks to pair, I use the general rules of thumb
Capped sources can be paired with fixed sinks.
Trickle sources can be paired with repeatable sinks.
For a fixed length game, investment sources always turn into capped or trickle sources. So you can use the previous two rules. You can be paired with exponential or competitive sinks. These are higher power and will sop up your sources. However they can be overkill, introducing mechanical complexity you may not need.
Balancing an ongoing game
For an ongoing game, you again set up your per chain spreadsheet of sources and sinks.
Since time keeps going, you don’t know how long a player will play.
One trick is to think in terms of balancing within a period of time. How do flows add up within a week or a month or season of play? If you can find that repeatable ‘long session’, you can model that out and find how sources and sinks balance.
When picking which sources and sinks to pair, I use the general rules of thumb
Capped sources can be paired with fixed sinks. These usually show up only during fixed-length sub-games within the live game. Examples include tutorials and limited-time events. You may want to avoid fixed sinks outside of these situations. Time is infinite and long-term players will overwhelm fixed sinks.
Trickle sources should be paired with strong repeatable, exponential or competitive sink.
A grind or investment source will always swamp a fixed or repeatable sink. Instead pair them with exponential or competitive sinks.
You almost never want a true exponential source. Long term, these make your life painful. They are hard to model mentally and small mistakes result in extreme resources pooling.
Try capping your investment sources. Any exponential power upgrades in an RPG are usually controlled with a hard level cap. Another common technique is to limit the number of investment slots you can use. Both these options turn an investment source into a more manageable trickle source.
Idle games pairing an exponential source with a slightly higher power exponential sink. Even these structures don’t last forever since exponential investment waits start becoming boring. So they rely on hard resets via ascension mechanism to escape the trap of their exponential sources.
Lean towards taut chains
When balancing sources and sinks, you typically want the sinks to be a little larger than the sources.
Not too much or that results in dead points in the spend where players suffer from painful scarcity.
Not too little because then you get pooling of resources and lack of pull on activity nodes within your value chain. The chain should always be pulled somewhat taut.
If you’ve done your work matching source and sink power and you’ve isolated your value chains (see multi-chain architecture below), the exact balance numbers are less important than you might imagine.
The emotions of scarcity and abundance
Once you get your general economy balance under control, there’s a huge amount of emotion you can extract from relatively minor variations in tautness.
Scarcity: When players feel scarcity, they’ll be highly motivated to search out and harvest scarce resources. They’ll experience anxiety and a tendency to horde.
Variability: When players feel high variation in availability, they experience similar emotions.
Abundance: However if you give them abundance, they’ll momentarily feel a sense of freedom and can invest in non-scarcity driven behaviors. Note this is different from the extreme overabundance mentioned above when we discussed imbalance economies.
Hedonic adaptation: However, if they experience abundance for too long, your value chains grow slack. Players stop finding meaning in earlier nodes and just rely on their pooled resources. The joy of abundance returns to a baseline.
The best games create an ebb and flow between scarcity and abundance within a narrowly controlled band of economic outcomes. A well balanced economy is a tool for driving rich player experiences.
You can play with these much like playing notes or pacing on a music instrument. For example, in Cozy Grove, we made harvesting very reliable. Most sources were capped daily so they acted as slow trickle sources. And in general, we leaned towards abundance. Not always. Various events or quests would suck up resources from the player’s hoarded supplies so abundance wasn’t 100% reliable. This helped combat hedonic adaptation. The result is a very low stress, cozy economic game pacing.
Note: Mimicking sink power by increasing sink magnitude
In a fixed length game, you can always approximate a higher power sink with a large magnitude low power sink.
For example, I have a game that lasts 10 turns. It has a trickle source of gold that produces 5 gold per turn.
Option A: I could pair the trickle source with a repeated gold sink that lets me buy 1 victory for every 10 gold. By the end of the game I’ll be able to purchase 5 victory points with no left over gold.
Option B: However, I could also create a single fixed sink that lets me purchase 5 victory points for 50 gold. I’ve taken a simple fixed sink and just increased the magnitude enough that it sops up my gold income.
On the surface, these look like the same end result. But they aren’t the same experience.
Large sinks can feel grindy since players have to put in a lot of effort before they can spend. There’s a music-like pacing to how players interact with economies. Beware of large gaps where players lose track of the tune.
If you end up changing the length of your game, you need to immediately go back in and rebalance all your sinks (or sources if you want to approach it from earlier in the chain.) In general, perfectly pairing sources and sinks of the same power that are balanced only by magnitude adds brittleness to your economic architecture.
In a long-term ongoing game, you cannot mimic a higher power sink with a much higher magnitude sink. In the long run, a higher power source will always swamp a lower power, yet high magnitude sink. Just the way the math works (feel free to graph it!)
In general, I try to avoid replacing sink power with sink magnitude. It is a bad habit to get into.
Issue: Content treadmills
Long term, heavy use of capped sources and sinks lead to content treadmills. A content treadmill is when you need repeated injections of new content to keep your Game-as-a-service (GaaS) running.
From an economic perspective, In order to extend the game, you need to add more sources and more matching sinks. Each of these requires a fixed amount of content. It can be better to invest in repeatable sinks.
Issue: Marginal value erodes over time with repeated actions
Even trickle sources with strong sinks can wear out. Imagine you get one apple a day and you eat one apple a day. Trickle source, repeatable sink. What is the value of the apple to the player on the second day?
In a simple model, the player has zero memory for the previous day. So they should look at the apple on the second day, realize they are hungry and be absolutely delighted to get a new apple.
Burnout: In practice, each new apple provides a decreasing psychological benefit. Players slowly get bored with yet another apple. Repetition matters in experiential goods. More of the same, even though it provides the same functional benefit will provide less novelty or mastery benefit.
Leverage: High leverage content are actions within your value chain that can be repeated many times without burnout. Most actions can only be repeated a small number of times before players get bored. The term ‘leverage’ comes from content that results in high ratio of gameplay relative to the cost of producing that content.
For example, the classic leverage on exploiting a weak point in a Nintendo boss is ‘three’. The first time you learn their weakness. The second time you practice exploiting the weakness. And the last time you demonstrate your mastery. But that action starts to get boring if you are asked to repeat it the fourth time because it is usually just rote pattern execution.
Solution – content recharging: The good news is that humans forget. An apple every single day might be low value. But if you let people forget about apples, and then give them an apple two months from now, that apple might again have high value. You can recharge content and regain some degree of leverage at the rate it takes for players to forget about that content.
CHAPTER 3 – ARCHITECTURE OF MULTIPLE VALUE CHAINS
Most games have multiple value chains. If you were to lay out all the value chains on a single piece of paper, you’d find that certain nodes are present in multiple chains. This creates a crisscrossing spaghetti of resource flows that is the complete value network for your game.
It is helpful to organize this ball of spaghetti in a fashion that is easy to understand and manipulate. Patterns for organizing your value network are your value architecture.
There are an infinite number of value architectures out there. But we want to focus on sorting our chains in ways that best satisfy the following goals
Independence: Each individual chain is easy to independently balance so that it doesn’t accidentally unbalance other chains.
Modularity: In GaaS, you want the option to easily retire old chains and add new ones as the game ages.
The most common architecture: Parallel value chains
The safest structure is to keep your value chains parallel to one another so they don’t overlap. Each set of action nodes is served by a set of unique resources and the player doesn’t need to make trade off between each chain. You can still feed multiple chains into the same motivational anchor as long as the anchor is multi-dimensional enough to be better satisfying by a little variety.
This mostly satisfies our goals
This lets you balance each chain in isolation.
It is easy to add a new value chain for an event and then remove it when the event is over. Or if there’s a piece content that players are burning out on, we can retire it without upsetting the balance of the rest of the game
Lots of bespoke resources. Each value chain needs its own resources that are not used in other value chains. As parallel chains multiply, so do resources and you’ll need a reasonable inventory system to track them all. This can add a lot of cognitive load for new players.
Fewer emergent interactions between systems: Since economic systems are isolated, you get fewer ‘interesting’ feedback loops. This is an intended outcome of the architecture, but worth acknowledging what you give up by adopting it.
Most long term GaaS evolve towards some flavor of architecture that contains multiple parallel value chains. It shows up again and again in various MMOs and F2P games. Players may not enjoy the explosion of currencies and resources that result, but they serve a real architectural need.
Architecture for applying buffs
A very common architectural structure folks build with value chains is to create a buff or boost that increases the efficiency or effectiveness of some other node in a different action chain.
In Jesse Schell’s terminology, these often take the form of ‘virtual skill’. A player purchases a +10 sword of smiting that boosts the amount of damage they can do to an enemy, thus allowing them to kill it faster.
Efficiency anchor: New designers often think that this virtual skill primarily serves a skill mastery anchor, like getting better at fighting in real life.
In practice, it serves as an efficiency motivation. (Or a power fantasy. The exact anchor depends in large part upon theming and audience) It has little to do with player learning and everything to do with a number being modified. The action in the economic node becomes cheaper to perform.
There are lots of possible efficiency boosts you can trivially build into your value chains. Basically any form of cost (time, money, resources, complexity) can be reduced by a boost.
Adding sinks to boosts: And you can add additional sinks into the Apply Boost node. For example, you might get a spell that increases damage output. But it requires mana to cast. Remember, every node is an opportunity to add another sink if you need one.
This is a very flexible and useful pattern that once you understand it, you’ll start seeing it everywhere.
Issue: Multiple undifferentiated inputs to node
Sometimes however, you have to cross value chains. There are helpful and unhelpful ways of doing this.
Consider the following unhelpful scenario that is unfortunately baked into most RPG systems. Here we have multiple nodes producing an undifferentiated resource.
In this example, we have two value chains that merge into one.
Value chain A: You can spend time and health killing Monster A
Value chain B: Or you spend those same resources killing Monster B.
In both cases, you get XP that you spend on leveling up. The value chain continues on after that, but we’ll just look at this snippet of the whole.
We see two things happen in this design pattern when a long term player groks the full value chain topology
First they realize they can make a choice. They can invest their time in killing either monster A or monster B.
Next they realize that if both monster A and monster B are plentiful, their time is always limited. So for efficient play, they should focus on killing the monster that gives the most XP for time spent. Let’s assume that’s Monster A.
As the player gains expertise, they’ll start to completely ignore Monster B. Even though it has book value, the marginal value comparison means that it is in practice valueless to the player.
This pattern has major implications on your economic design. In MMOs you may create 100s of enemies. Or hundreds of raids or quests. Yet, players will insist on playing only one or two. All that content you spend so much time and money developing is essentially wasted.
This structure is very difficult to balance since players only care if the two sources are perfectly equal. And it never is. Even in cases of mathematical equality, there are cultural, habitual or aesthetic factors that cause players to prefer one path over another. This architectural decision ends up invalidating big swathes of content.
Cap each source: If there are a limited number of times you can exercise Source A and also a cap on Source B and you need to to engage with both in order to satisfy the subsequent node. This is the most common answer for single player games. Here you have a fixed budget of content and can plan out exactly how much players should consume before they unlock the next elements.
Multiple currencies: For more complex economies, it can be far more robust to use multiple differentiated input resources. The next section goes into more detail.
Pattern: Multiple differentiated inputs to node
Now let’s consider an alternative topology
Same as before you spend time and health killing Monster A. But this time, you get a unique resource, horns. And monster B gives gems. And in order to level up, you need both Horns AND Gems.
This setup has a very different set of player choices
The player must engage with both Monster A and Monster B to level up. If they only kill Monster A, they’ll lack Gems. If they only kill Monster B, they’ll lack Horns.
The level up node creates a strong pull on the subsequent action nodes, giving these actions clear value.
Players can choose the order that they engage with Monster A or Monster B, but they cannot ignore them. If there’s substantial content associated with those earlier nodes, you guarantee that it will be seen as valuable and that players are incentivized to exercise it.
Pattern: Overflow from one chain to another
Suppose you want a player to pursue one value chain for a while and then switch over to a different value chain later in the game’s progression.
In this example
Killing a monster gives gems
Players can spend gems to level up.
However leveling up is capped at level 10
Once players finish leveling, they can pour excess gems into crafting decorations.
This overflow pattern is useful when you have fixed sinks. You set up cascading pools so that when one is filled, the excess can flow into others.
This can be a useful pattern as well if your game is serving multiple player motivations. Which is almost always the case since any sufficiently large player population will contain multiple playstyles driven by multiple motivation.
Say you have some players who love to decorate and others who like to progress.
You want both to keep performing the core loop of killing monsters.
So you use this structure to pull gems from the core activity, but then give them a choice on how they want to spend their hard won resources. Each path is anchored on a different motivation.
Pattern: Lock-and-key choices
Earlier we covered how a single currency can lead to choices where players pick the most efficient path and ignore the rest. In a large, loosely controlled economy, this can cause major balance issues. However, there are more controlled variants where the player’s choice of how to spend esources are the most interesting part of the game.
The common elements of this pattern include
Key resource: There’s a capped source producing a resource. This is the metaphorical ‘key’ in a ‘lock-and-key’ node-resource pair.
Lock node: Gated nodes are unlocked with key resources.
Choice: There are always more available lock nodes than key resources, so players need to make clear choices about which option to invest in.
Opportunity costs: By selecting a node to invest in, you lose the option of gaining resources from the other lock nodes.
Common examples of this
Worker placement: The player gains access to a very limited number of workers. Those workers may be assigned to limited jobs to produce other resources or buffs. Or other workers! Sometimes there is a cost to place the worker. Or a cost to remove the worker. But critically, there are never enough workers to fill all the possible production stations so choices must be made.
Skill trees: The player gains access to a very limited number of skill points. Those points are assigned to unlock skills in a predictable skill tree. This creates both a buff for the player and opens up the chance to unlock future skills further down the tree.
Designers have a lot of control using this value chain pattern. They can change the benefits of each lock node and balance them against one another. They can control how many choices are valid by altering the amount of key resources. Content is invalidated when players make a choice, but the amount and impact of that is up to the designer.
New designers often mistake multiple undifferentiated inputs as the serving the same role as lock-and-key choices, but once you know the structure of the value chains, you can see they are quite different. Lock-and-key choices always ensure a strong pull (and a taut chain) while multiple undifferentiated inputs result in dangling chains that are left unexercised.
CHAPTER 4 – ENDOGENOUS VALUE NETWORKS
“According to the dictionary, one definition of endogenous is “caused by factors inside the organism or system.” Just so. A game’s structure creates its own meanings. The meaning grows out of the structure; it is caused by the structure; it is endogenous to the structure.”
This quote has stayed with me for almost two decades. Value chains are a method of formalizing this fundamental truth into a useful design tool. They start to get at the heart of how meaning is constructed within a game.
Value modeled as value networks
“Meaning” and “Value”are vague terms that we need to define more clearly in order to design . Value chains model “value” in terms of common elements of a game (actions, resources) arranged in a network topology. This allows us to get far more explicit about what value we are designing into our game. We gain visible levers and knobs we can manipulate.
Value networks are internally self-supporting
Most elements in a game have value due to their relative relationships with other elements in the game.
If you take away the other elements earlier or later in the value chain, the game loses meaning
Change the balance or nature of the relationship between elements, the game loses its meaning.
Games as artificial spaces
Most game value networks are artificial. They are arbitrary and cut off from reality. This artificial space is often called the ‘magic circle’ within which gameplay exists. This artificiality provides such creative freedom! We are building cartoon worlds that don’t need to mimic the difficult-to-work-with structures found in natural economies.
For example, when designing a giraffe refuge in the natural world.
Does anyone even want a giraffe refuge? How are you going to pay for it?
Then designers need to take into account years of law, history, logistical issues associated with limited physical space, connections to adjacent spaces, and whether or not your neighbor is allergic to giraffes.
There are an immensity of constraints and unexpected feedback loops that are impossible to fully capture in any simplified model.
None of those rules apply in a game about building a giraffe refuge.
We can set up artificial rules where giraffes are plentiful, land is plentiful and everyone loves giraffes.
We can create grokkable linear value chains and eliminate undesired feedback loops.
We can intentionally design an artificial world where it is easier to build playful giraffe-centric activities within.
Ultimately a game’s magic circle is anchored in reality
The “magic circle” is the conceptual boundary where a player opts into the value structure inside a game. Players opt into the magic circle of a game by saying “You know what? I know this virtual stick isn’t real. But I’m going to play along and act as if it has meaning.”
But in the end, we should never forget that the reason why the player participates in the game’s value network is because they are seeking real-world value. This is why every value chain ends with a motivational anchor. Personal needs fulfillment always pierces the boundary of the magical circle.
Unmet player needs: Play is a seeking behavior. You have unmet needs, but you don’t know how to fill them. So you experiment in a safe fashion to understand your options. This last step is the definition of play.
Game makes player promise: A new game makes a promise to the player, usually rooted in the meeting of some need. Diablo promises power and mastery. Animal Crossing promises a relaxing respite. World of Warcraft promises mastery and friendship. This is the hook that gets you sucked into a game.
Onboarding: And there’s a grace period. Because the point of play is to wander about for a bit and figure out how to meet your needs. Even players know that need fulfillment can’t happen immediately. Mastery can take many hours. Social bonding can take weeks. Players need to build up the tools. They need to understand the path forward. So players willingly run through tutorials. They willingly follow the chain of quests. Onboarding runs on goodwill that their needs will be eventually met. This step is introducing players to the early stages of the value chain.
Understanding the path towards: That goodwill runs out. As soon as the promise is made, a timer is ticking and the player is thinking in the back of their mind “How is this game going to fulfill its promise?” The job of the game is to paint that path. And demonstrate real progression towards it. If the game doesn’t help the player understand how all this (expensive) playful activity will ultimately fulfill a key motivational drive, they will stop playing. The game must connect the dots. This is making the value chain visible to the player.
Demonstrating need fulfillment: In as short a timeframe as possible, the game should provide player experiences that fulfill the needs as they were promised. This is the end anchor of the value chain.
So all of our elaborate value scaffolding does need to serve the player’s needs in the end. Every cartoon, hyper-designed endogenous game system contains a connection to the real world. Because games are played by real humans with real human needs.
Value chains should give you a strong framework for planning and balancing your game economy. You’ll be able pinpoint issues and communicate targeted balance fixes using a common language. The technique targets faucet-and-drain economy designs, but since this remains the dominant method used across most popular genres, you should be well equipped.
Game economy design is a much richer topic of technique and practice than I could possibly cover in this paper. Many modern designers find they devote years of their career to learning the nuances specific to their genre and their community’s needs. If you are interested further in this topic, I highly recommend the following:
Breakdowns: Take one of your favorite games. Identify the individual value chains. Be sure to include the anchors! Make notes on the architectural elements such as branching or choice built into the chains. Ask yourself what you could have done differently to serve your unmet needs better. Also do this exercise with one of your least favorite games.
Game jams: Very few large teams will give an unproven designer the responsibility to design an economy from scratch. However, many of the fundamentals can be practiced on smaller game jam-sized projects. Limit the number of length of your value chains. But try them out! Try out strange new architectures. Playtest! Balance these tiny games. The lessons you learn scale to larger projects.
There are also many further areas of investigation for those interested in extending value chains as a design tool.
Trade: How do value chains map to more open economies with features like player-to-player trade?
Visualization: Is there value in reconstituting value chains into a more traditional spaghetti diagram? Such visualization tools don’t yet exist. But you should be able to composite value chains together automatically and perhaps even summarize them.
Ethics: Can we use economy design for good? The use of value anchors deliberately centers human needs as the primary driver of value. Yet the world is rife with reductive, selfish ideologies that flatten the richness of humanity to mere numbers (homo economicus, libertarianism, much of current crypto.) Economy design is an amoral tool. It requires ethics, compassion and a keen eye for spotting externalities in order to avoid causing immense systemic harm.
(Note: This essay was originally published on Gamasutra.com in 2007. To help prevent link rot, I’m reposting it here with minor edits and fixed links. I’ve also included a workshop presentation on the topic if folks want to teach these ideas in their classes.)
1. Moving Beyond Alchemy
“…it was clear to the alchemists that “something” was generally being conserved in chemical processes, even in the most dramatic changes of physical state and appearance; that is, that substances contained some “principles” that could be hidden under many outer forms, and revealed by proper manipulation.”
I recently happened across a description of alchemy, that delightful pseudo-science of the last millennium that evolved into modern chemistry. For a moment I thought the authors were describing the current state of the art in game design.
Every time I sit down with a finely crafted title such as Tetris or Super Mario Brothers, I catch hints of a concise and clearly defined structure behind the gameplay. It is my belief that a highly mechanical and predictable heart, built on the foundation of basic human psychology, beats at the core of every single successful game.
What would happen if we codified those systems and turned them into a practical technique for designing games?
In a Time Before Science
“Throughout the history of the discipline, alchemists struggled to understand the nature of these principles, and find some order and sense in the results of their chemical experiments—which were often undermined by impure or poorly characterized reagents, the lack of quantitative measurements, and confusing and inconsistent nomenclature.”
Historically, the process of understanding games has been limited by numerous factors ranging from messy experimental practices, spiritual reliance on untested theories of play, and confused terminology. We are still alchemists of our trade, mixing two-parts impure story with one-part polluted game play with three-parts market voodoo.
As an industry, we need to go beyond the mystical hand waving that defines modern game design. It is now possible to craft, test and refine practical models of game design built from observable patterns of play. We can describe what the player does and how the game reacts. Recently, we’ve begun to crack open why players react to certain stimuli and are able to create models that predict pleasure and frustration.
This essay will describe one such model.
Fundamental Science Forms The Future
Diagram 2: Condensation polymerization of Nylon, (a substance not available to alchemists)
The bigger hope is to move our alchemical craft towards the founding of a science of game design. We currently build games through habit, guesswork and slavish devotion to pre-existing form. Building a testable model of game mechanics opens up new opportunities for game balancing, original game design and the broader application of game design to other fields.
The advent of basic chemistry gave us tools to build a new world of technologies far beyond that imagined by our alchemist forefathers. Plastics, engines, fabrics, power sources revolutionized our lives. It is a worthy effort to crack the fundamental scientific principles behind the creation of games.
2. The Foundations Of A Model Of Game Design
Where chemistry separated itself from alchemy by building testable models of physical atoms, a science of game design concerns itself with testable models of human psychology.
Many of the attempts to define games have focused on the mechanistic elements of the game, such as the primitive actions that the system allows the player to perform or the tokens that the player manipulates. The approach has been to treat games as self contained logical systems.
Mechanics and aesthetics are certainly important pieces of any model of game design, but in the end, such analysis provides little insight into what makes a game enjoyable. You end up with a set of fragmented pieces that tell you almost nothing about the meaningful interactions between the game as a simulation and the player as an active and evolving participant. Games are not mathematical systems.They are systems that always have a human being, full of desires, excitement and immense cleverness, sitting smack dab in the center. To accurately describe games, we need a working psychological model of the player.
Our player model is simple: The player is an entity that is driven, consciously or subconsciously, to learn new skills high in perceived value. They gain pleasure from successfully acquiring skills.
Diagram 3: The player follows clues to the acquisition of a new skill
Let’s dig into three key concepts in our player model.
Driven to learn
A skill is a behavior that the player uses to manipulate the world. Some skills are conceptual, such as navigating a map while others are quite physical, such as pounding in a nail with a hammer. This ties into an intrinsic motivation towards self-determination. “I want to do what I want to do. And skills help me get there.”
Driven To Learn
Play is instinctual. In low stimulation environments where we are not actively pursuing activities related to food and shelter, people will begin playing by default. Strong feedback mechanisms in the form boredom or frustration prod us into action. Given a spare moment, we throw ourselves into playing with blocks or dolls as children and more intricate hobbies as adults. It is a sign of our need for meaningful stimulation that solitary confinement remains a vicious punishment for the most hardened criminals.
The flip side is that we are rewarded for learning. The sensation that gamers term ‘fun’ is derived from the act of mastering knowledge, skills and tools. When you learn something new, when you understand it so fully you can use that knowledge to manipulate your environment for the better, you experience joy.
There is a reasonable amount of neuroscience available to support this claim. Edward A Vessel, a cognitive neuroscientist at the NYU Center for Neural Science writes:
“These “aha” moments, when a concept or message is fully interpreted and understood, lead to a flood of chemicals in the brain and body that we experience as pleasurable. It feels good to “get” it. The deeper the concept is, the better it feels when we are finally able to wrap our head around it.”
Upon the click of comprehension, a natural opiate called endomorphin, a messaging chemical in the brain similar in structure to morphine, is released. As humans, we are wired to crave new information constantly. In some sense, what you and I term curiosity can be interpreted as our brain looking for its next fix of deliciously fascinating information.
As game designers, we deal with the fun, boredom and frustration on a regular basis. It is good to recognize that these are biological phenomena, not some mystical or mysterious sensation. For more thoughts on the topic, I encourage you to have a quick read through Raph Koster’s book “A Theory of Fun for Game Design”
Players pursue skills with high perceived value over skills with low perceived value
Play is, perhaps counter intuitively, a deeply pragmatic activity. Our impulses to engage in play are instinctual, selected for by evolution because it provides us with the safe opportunity to learn behaviors that improve our lot in life without the threat of life threatening failure. We play because we are built to expect the eventual harvesting of utility from our apparently useless actions. We stop playing when we fail to find that utility.
The perception of value is more important than an objective measurement value. Humans are not creatures of pure logic. We know people exhibit consistent biases in how they weigh their actions. For example, they’ll often undertake bizarre risks because they are unable to properly evaluate statistical odds. We’ve also realized that people have substantial limits on how much information they can take into account when making any one decision. Many decisions are made based on highly predictable ‘gut’ reactions that have their own subconscious rules.
3: Interaction Loops
With our player model in hand, we can describe how the player interacts with the game.
The basic ingredients of a game are, if not standardized, at least well described in a variety of books and rambling by designers across the past decade or two. I’ve taken the basic ingredients of tokens, verbs, rules, aesthetics, etc and remixed them into a self contained atomic feedback loop called an interaction loop. Each unit describes how the player gains a new skill.
Diagram 4: The player follows clues to the acquisition of a new skill
An interaction loop feedback loop is composed of five main elements:
Decision: The player observes any known affordances and weighs possible outcomes.
Action:The player performs an action. For an interaction loop encounter by a new player, the action might involve pressing a button. More advanced atoms might instead require the player execute a batched set of actions such as navigating a complex maze.
Simulation: Based off the action, an ongoing simulation is updated. A door might open.
Feedback:The game provides some form of feedback to the player to let them know how the simulation has changed state. This feedback can be auditory, visual, or tactile. It can be visceral in the form of an exploding corpse or it can be symbolic in the form of a block of text.
Modeling: As the final step, the player absorbs the feedback and updates their mental models on the success of their action. If they feel that they have made progress, they feel pleasure. If they master a new skill or other tool, they experience an even greater burst of joy. If they feel that their action has been in vain, they feel boredom or frustration.
A shorthand diagram that I find useful for recording atoms is as follows:
Diagram 5: Our canonical interaction loop
For example, let’s dissect the act of jumping in Mario
Diagram 6: The interaction loop of the player learning how to make Mario jump
Decision: A player notices a button on the controller. They know from past experience, it is pressable.
Action: An inexperienced player pushes a button.
Simulation: The simulation notes the action and starts the avatar of Mario on the screen moving in an arc.
Feedback: The screen shows the user an animation of Mario jumping.
Modeling: The user forms a causal mental model that pressing the button results in jumping.
Implicit in this model is that the atom is often looped through multiple times before the user understands what it teaches. The first pass may only clue the user that something vaguely interesting happened. The user then presses the button again to test their theory and Mario once again bounces up into the air. At this point, the player smiles since they realize they’ve acquired an interesting skill that may be of use later on.
This Thing We Call Play
“Man is a Tool-using Animal … Nowhere do you find him without Tools; without Tools he is nothing, with Tools he is all.” – 19th century essayist Thomas Carlyle
Upon the acquisition of a shiny new skill from a skill atom, players experiment with it. They try it out in different environments and see if it does anything useful. This semi-random exploration is the classic ‘play’ activity that we see children perform. For example, when a new player masters how to jump, you’ll notice they’ll almost immediately start happily hopping about the level. On the surface, it is a silly frivolous activity. In reality, we are observing humanity’s instinctual process of learning in action.
In the course of experimenting, the player will occasionally stumble across something in the environment that gives them interesting information that might lead to the mastery of a new skill. At this point, you’ll see the behavior of the player become more deliberate. A mental model begins coalescing in their minds. In our jumping example, the player starts bumping against a platform. They may even reach the top of a platform. It is very common that skills acquisition requires multiple passes through the new skill atom before mastery is achieved.
Eventually, the player uses an existing skill to grok another skill. They experience a wash of pleasure and start the process all over again.
Chaining Of Game Mechanics
We can visually represent how players learn by linking our basic interaction loops together to create a directed graph of atoms called a skill chain.
Diagram 7: Two linked atoms
The skill from one atom feeds into the actions of another atom further down the chain. By linking more and more atoms in, you build a network that describes the entire game. Every expected skill, every successful action, every predicted outcome of a simulation, every bit of required feedback can be included in a simple, yet functional fashion.
A skill chain is a general notation that can be used to model pretty much any game imaginable. Your design can be broken down into dozens of simple atoms that link together to form a clear and easily readable map of how the game plays. The skill chain, with its ability to describe the player experience instead of the mere mechanics of the game, provides a far richer description of the meaningful moments that occur during gameplay.
How Players Interact With A Skill Chain
Players will travel from atom to atom like Pac-Man following a trail of dots towards the power pellet. They move from one skill to the next even when they have only a vague concept of the ultimate destination.Chomping up those dots is good.
One of our peculiarly human limitations comes into play at this point. Players are unable to predict the value of a new skill more than a couple atoms down the chain. As long as there is a new skill with potential value within our prediction horizon, players will pursue it. There may be no actual long term payoff other than the pleasure of the experience, but we don’t care. As long as there is a promise of a long term payoff and the short term rewards keep coming, we assume that there will be some final benefit from our efforts.
Diagram 9: Players have limited foresight
If you look at this from an evolutionary perspective, our behavior makes quite a bit of sense. Many useful skills take upwards of five to 10 years to master. During those early days of our education, the basic playful activities such as gossiping about which kids have cooties seem rather silly. Later on however, our mastery of politics, science, or in the case of the cooties, mating rituals, yields a hugely positive impact on our well being.
The just-so story here is that playful folks that instinctually engaged in long term learning with no immediate benefit were the ones that mastered agriculture, hunting and language. These folks thrived. Those that did not died off.
However, our brains never evolved to deal with modern games. The existence of a set of interaction loops that are tuned just to entertain us and that never actually lead up to a real world skill is something new to the world. At their most puerile, games are a grand hack. The minute by minute experience fits all our biological heuristics and sounds all the right bells. So we keep on playing. And we wonder why so many games have such horrible endings.
4. Status Of Atoms In The Skill Chain
A skill chain provides some rather useful information about the state of the player as they engage the game. Imagine that the skill chain is the instrumented dashboard that lights up with the player’s progress. At any point in time you can tell the following information
Mastered skills: Skills that have been recently mastered.
Partially mastered skills: Skills that the player is toying with, but has not yet mastered.
Unexercised skills: Skills the player has yet to attempt.
Active skills: Skills that the player is actively using. (aka the Grind)
Burned out skills: interaction loops that the player has lost interest in exercising.
Diagram 10: Icons for skill status
We’ve talked a little bit about mastered and partially mastered skills. Unexercised skills are pretty self explanatory. If a player can’t perform the actions necessary to understand a skill, that atom will never be exercised or mastered. Mastery flows down the chain and if players are blocked early on, they’ll never reach the further atoms.
The two states that are worth a bit more explanation are active skills and burned out skills.
The player only experiences the joy of mastery for an atom only once. After the moment of mastery, a biological feedback system kicks in that dampens the pleasure response to exercising those same pathways again. What was once exciting becomes boring.
However, players will continue exercising an already mastered atom as a new tool for manipulating their world. A mastered atom is as good as a shiny new hammer hanging from a workman’s belt. When a new opportunity comes up, typically in the form of an atom further down the skill chain, the player makes use of their new skill to advance their knowledge.
Players have enormous patience. They are willing to exercise a basic interaction loop thousands of times in order to achieve mastery of a higher order atom. Players jump innumerable times in Super Mario Brothers in order to reach more powerful skill sets further down the chain.
A skill that has been mastered and is now simply being used to activate other icons is represented by the lit light icon.
Diagram 11: Active Icon
Players don’t always bridge the gap between one atom and the next. They master a new skill, they play with it but fail to find any interesting use for it. This is known as burnout.
Diagram 12: Burned out icon
For example, suppose our player pressed the jump button. They performed the jump and we recorded their mastery of the skill. However, this particular player never figured out how the jump might be useful. Perhaps they didn’t jump near the platform and receive interesting feedback on the next atom.After a short period of experimentation with no interesting results, the player stopped pressing the jump button entirely.
When a player burns out on a particular atom, the consequences ripples up and down the chain.
Early Stage Burnout
In the example above, the Reach Platform atom will never be mastered. The foundational skills are not in place. In a deeply linked skill chain, a burnout early on can chop off huge sections of the player’s potential experience. You can think of learning curves in terms of managing early stage burnout.
Later Stage Burnout
On the other hand, a burnout later on down the chain can devalue active skills.
For example, assume we have a single platform in our jumping game and there is really nothing on it. The player jumps on the platform, discovers no interesting new activities and so stops jumping on platforms. This, in turn, atrophies the Jump skill, because if the player doesn’t need to jump on platforms, why would he bother jumping?
Burnout Is Our Gateway To Testability
Burnout is a very clear signal that our game design is failing to keep the players attention. As you watch burnout creep across a game’s skill chain, it is a signal that players will soon stop playing the game.They are becoming bored, frustrated and perhaps even angry.
Perhaps most importantly, we can measure when burnout occurs for an individual atom. This gives us, as game designers, unprecedented qualitative insight into how a particular design is performing with play testers. When you start tracking burnout along with the other skill states, you can visualize the problematic areas with great clarity and accuracy. The entire topic of measuring performance of a game through instrumentation of its skill chain is a rich topic for further exploration.
Diagram 13: Skill atrophy due to later stage burnout
5. Advanced Elements Of A Skill Chain
We’ve covered the basic elements of a skill chain and how to record that status of the player’s progress.There are only a few more pieces we need so that you can start building your own skill chains.
Pre-existing skills: How the skill chain is jump started.
Evocative Stimuli: How we represent story and other aesthetic aspects of modern game design.
Players bring an initial set of skills to a game. These skills always form the starting nodes of a skill chain. Accurately predicting the player’s existing skill set has a big impact on the player’s enjoyment of the rest of the game.
Diagram 14: How pre-existing skill feed into initial interaction loops
Lack Of The Correct Initial Skills
If the player lacks expected skills, they will be unable to engage the initial atoms in the game. In our example about jumping, imagine a player that didn’t realize that you need to push the button on the joystick in order to do something. Such an example may seem ludicrous, but it is one faced by many non-gamers whenever they are faced with a freakishly complex modern controller. Many game designs automatically assume the ability to navigate a 3D space using two fiddly little analog stick and a plethora of obscure buttons. Users without this skill give up in frustration without ever seeing the vast majority of the content.
It is very important to realize that such users aren’t stupid. They merely have a different initial skill set. One of our jobs as designers is to ensure that the people who play our game are able to master the game’s early interaction loops. Ultimately this means making an accurate list of pre-existing skills for the target demographic and building our early experience around those skills. Don’t assume skills that may not be there.
Pre-mastery Of Skills Taught In The Game
The flip side of all this is that if players have already mastered existing skills, the process of mastering early atoms is likely to be quite boring. When a player, who has completed a dozen hardcore titles, plays a game sporting a 10-minutes navigational tutorial they become bored. All the reward notes are sour because their jaded brain doesn’t react at the appropriate points. If a game doesn’t teach the player anything new, the player is very likely to experience burnout on the early atoms.
Targeting the correct set pre-existing skills is a balancing act. If you choose correctly, you’ll end up with an ‘intuitive’ game that players enjoy. If you choose incorrectly, you risk frustration, boredom and inevitable burnout.
Evocative Stimuli using Arcs
Games are laden with story, setting, and imagery intended to evoke a particular mood and other intriguing but mostly non-functional elements. Gamers derive great pleasure from this feedback. We can represent much of this mélange of artistry with the use of a special type of atom known as an arc.
Arcs are atoms that the designer knows will never result in a useful in-game skill, but that still evokes the past experiences or mental schema. When the player experiences the information cues, existing player memories are activated and the brain greedily sucks up the clues. For example, many players have pre-existing associations with mushrooms. If you are of a certain age and a certain liberal background, you may even own a rainbow colored T-shirt that sports a mushroom or two. When such a person plays Super Mario Brothers for the first time, they are quite likely to perk up at the sight of magic mushrooms. An interaction loop in their brain is activated, they start activating ideas about mushrooms, and begin free associating why might dear Miyamoto have placed such a counter culture reference in the game.
Of course, the reality is that for the psychedelically minded, the mushroom imagery is flavor only.
Now these evocative arcs can be useful! If the player had read Alice in Wonderland, they might associate mushrooms with changing in size. In this case, the fact the mushroom makes you bigger already has a pre-existing mental pathway and when the player experiences it again, they are essentially reinforcing that path. So evocative arcs can influence what mental schema (existing skills) players tap into when forming models of cause and effect.
The downside of evocative stimuli is that most players rapidly burnout on such sleights of hand. The first time you see the mushroom, you might think it’s ‘mushroom-y-ness’ interesting. The second time, you see it as its utilitarian nature: An icon representing access to a tool (growing larger) that helps you navigate the world more efficiently.
We’ve covered a lot of ground in this essay. Hopefully, the diagrams give you a good understanding of how to describe a game using skill chains.
Using Skill Chains
As a tool, I’ve found that skill chain diagrams dramatically improve my understanding of how a game works, where it fails and where there are clear opportunities for improvement.
Creating a skill chain provides you with the following information:
Clearly identify the pre-existing skills that the player needs to begin the game
Clearly identify the skills that the player needs to complete the game
Identify which skills need feedback mechanisms.
Identify where the player experiences pleasure in your game
Alert the team when and where players are experiencing burnout during play
Provide a conceptual framework for analyzing why players are experiencing burnout.
Though it takes a little practice, interaction loops aren’t all that complicated to define and are really no more of a burden than writing unit tests for a chunk of code.
Skill chains are a deep topic and we’ve described only the most basics aspects of how they function.Further topics of inquire include:
I like to imagine that models like skill chains will help raise the level of intent and predictability in modern game design. With the concepts in this essay, you can start integrating this model into your current games and collecting your own data. We’ve got some immensely bright people in our little market and it is almost certain that they can improve upon this foundational starting point. By sharing what you’ve learned, we can begin to improve our models of design. What happens if game designers embrace the scientific process and start to build a science of game design?
The alchemists of ages past dreamt of turning lead into gold. They performed mad experiments with imprecise equipment and questionable theories of how the universe worked. Modern game designers are not really so different. Those not simply here for the sake of profit instead rally around equally fantastical dreams such as creating a game that makes the user cry or enlightening the world with games of politics or hunger. We crib cryptic notes from past successes and chortle merrily when our haphazard experiments manage to mildly entertain our audience. We are on the leading cusp of deep human / software interaction and yet we know so little.
It is only by gaining a deeper understanding of the fundamental building blocks of design that game designers will gain the power to break free from the accidental successes of the past. With practical techniques gained from controlled experiments, we will create radically effective new applications. When we have our basic chemistry, our basic systems of measurement and our basic atomic theory, perhaps then we can consistently build games that tap into the heart of human psychology.
The reproducible application of psychological manipulation of individuals and groups using software is big heady stuff. In the short term, I would hope that a deep understanding of models like skill chains help us crack open the rigid craftsmanship of existing genres so that we can build better, more potent games. Long term, it will be interesting to see what world changing uses we can find for our ever improving psychological technology.
References And Notes
Workshop on using interaction loops and skill chains
Irving Biederman and Edward Vessel, American Scientist, May-June 2006
Abstract: “From hand-held DVD players to hundred-inch plasma screens, much of today’s technology is driven by the human appetite for pleasure through visual and auditory stimulation. What creates this appetite? Neuropsychologists have found that visual input activates receptors in the parts of the brain associated with pleasure and reward, and that the brain associates new images with old while also responding strongly to new ones. Using functional MRI imaging and other findings, they are exploring how human beings are “infovores” whose brains love to learn. Children may enjoy Sesame Street’s fast pace because they get a “click of comprehension” from each brief scene.”
Relationship of Skill Chains to MDA (Mechanics, Dynamics, Aesthetics)
This is a question that has been posed on occasion. MDA is a game analysis framework put forth by Robin Hunicke, Marc LeBlanc and Robert Zubek. It is one of many descriptive techniques that categorize the elements of a game. MDA is particularly useful to new design students because it has the key insight that the player experience (what they call Aesthetics) is a second order effect derived from playing the rules of the game.
The major difference between the two approaches is that MDA stops there. There is little attempt to model how rules and feedback produce the actual player experience with the game. There’s just these fluffy, conceptual categorical buckets. Since there’s no casualty, MDA analysis also fails to provide any objectively testable structure. With skill chains, you can always hook up logging software and observe where atoms light up and where they burn out.
As game budgets expand once more, the success of a title often depends on producing large amounts of high quality content. This is not a trivial task. Mistakes setting up your content plans can easily result in panic, shipping delays, scope cuts, rework and crunch. Modern developers live on the content treadmill so we might as well embrace it.
For a long time I’ve been interested in content architectures, the tooling and data structures behind what content we make. This somewhat obscure topic drives much of the production efficiencies available to a team. A poor content architecture can easily result in an equivalent player experience costing 10 times as much time and labor. That’s the difference in output between a 30 person team and a 300 person team; a lot of money and human life to naively misspend.
Who this is for
Producers: Anyone above level of associate producer should know this topic down cold. To paraphrase the words of designer Crystin Cox, “I want to be able to ask a producer whether I should use a placeholder or a vertical slice when building an experience.” To a large degree this is your job since these early decisions drive much the team’s ability to deliver on a schedule and adapt to unexpected changes.
Designers: If you decide what the team makes, you owe it to them to also understand the best possible methods of building the desired outcome. Design leaders maximize the impact of the experience they deliver while working within a fixed budget.
Engineers: You’ll be building many of these tools and pipelines. Wouldn’t it be nice if they were useful? Wouldn’t it be nice if other disciplines could communicate their needs? Knowing how to think about serving content authors improves the game, your work and results in happier cross team relationships.
What we’ll cover
Content architectures are a broad topic best approached holistically. Existing content architecture experts are usually veteran developers who have multiple games and dozens of failures under their belt. Unfortunately that means this essay needs to spend time intro-ing the basics before we get to the more advanced considerations. Apologies for the slow build!
Terminology: Basic definitions of what we are manipulating in a content archichitecture.
Concepts: Key concepts that help us think about our content.
Constraints: What are the specific content choices for a given project that shape our architecture?
Basic architectural patterns: How might we organize our content?
Advanced patterns – Manual composition: How do we manage rigidities in the content pipeline?
Advanced patterns – Automated composition: How do we reduce rigidities with automation?
Meta – Tool authoring: How do we build tools that multiply our authoring efforts?
Let’s start off with some basic definitions that work for most forms of content you’ll run into. I’m abstracting the discussion away from specific content (levels, character, textures) so we have the building blocks to think conceptually about any content in our game. We want to get to “content algebra”, instead of always asking “how many apples does Bob have?”
Content: Content is an authored set of data intended to be displayed in some broader game system and consumed by the player to create a meaningful experience. More traditional forms of content include things like a chapter of a book, or a painting in a museum.
We often think of game content as generic media like 3D models or text. And that’s definitely where we spend a lot of effort. However each game also contains data files like a loot table, level progression or powerup. These need to be designed with care.
Chunks: Traditionally content comes in the form of content chunks. This is a piece of the player experience that is standardized and reproducible. Examples of chunks include
Level: A game like Super Mario Bros has discrete levels. Each module is self contained and consists of a set of platforms, enemies and win conditions. A game is then composed of multiple levels.
Player character skin: A bundle composed of 3D model, UVs, textures, shaders, animation rigs and state machines. The player has a choice between multiple skins.
Weapon: A set of properties for weapon damage, rate, cooldown. As well as associated art, economics costs, etc.
Player buff: A set of modifications that occur to an external set of properties. Along with constraints on when and how the buff is triggered.
The contents of each chunk differ. Weapon A has different data than Weapon B. But the data structure and how that data feeds into other systems is shared.
Standards: Standards are rules and constraints that define a content chunk. They help reduce risk by removing unexpected variability and associated thrash. They help improve quality by focusing an author on excelling at a particular well-defined space. They help improve efficiency by eliminating common blockers and streamlining workflow. They help teams scale, by allowing multiple authors to work coherently on the same project.
Sets: These chunks are organized into sets. You might have 20 levels in a game. That’s your game’s set of levels. Or 500 barks. That’s your game’s set of barks.
Composition: Chunks can be assembled together into new composite chunks. A level chunk is a composite of enemies, level tilesets, powerups and other modular elements. The level designer likely did not create any of these sub-components, but they put them together to form a unique player experience.
Composition is a creative act of authoring. Someone needs to make deliberate choices on what is included and its relations with the other elements. Even a writer composes words they did not create on the page. A painter composes color they did not create on a canvas.
Dependencies: When you split content up into chunks and string them together in a content architecture, we create dependencies. In order for content to work or have meaning, it requires that other content or systems are functioning exactly as expected. The act of creating chunks always creates dependencies since there’s a fuzzy line for where content wants to reside. Standards help catalog and isolate dependencies. Later on, we’ll see many of the tools for managing content architectures are about structuring dependencies in a useful manner.
Questions worth asking about your game: But we often don’t take the time to think of ‘words’ or ‘color’ as standardized chunks. They are just the invisible air we breathe. Eliminating our blindness to the ‘intentionality of the default’ is the first step one must take. As a content architect you need to expand your perspective and see these elements as explicit design choices.
What are your chunks?
What are their standards?
What are their sets?
How are they composed?
What are their dependencies?
In order to design a content system, it helps to have a mental model of how content ‘works’. Here are some of the big picture rules of content authoring.
Content delivers value: We author and deliver works of art to players in order to provide them with meaningful experiences. We can build content that harms or wastes a player’s life. Or we can build content that enriches their life.
Content is consumed: Content can be experienced a certain number of times before players feel like they understand the experience and are ready for something different. Some content becomes a touchstone for an ongoing socio-economic player ritual, but most is used and then put aside. The player exhausts their motivation to return to the content.
Consumption is iterative. Players experience a chunk of content 1 to N times before they move on. Chunks that are experienced once and then discards are seen as Highly Consumable. Ones that can be experienced many times without being discarded are seen as Evergreen.
Authoring is iterative. How does an author deal with the uncertainty inherent in a diverse audience’s consumption of the content? You iterate. You deploy the content and observe the reaction of those consuming it. Then you revise the content and test once again.
At the most basic level, authors do this with themselves in a process called ‘self playtesting’. They switch between a creation state and a consumption state. With writing and painting this happens moment-by-moment in a tight iterative loop. For example when writing, the following happens thousands over times:
I write a word
Then immediately read what I wrote and react.
Then I revise.
Games have longer feedback loops than many forms of media. As we author, we can imagine in our minds how it might play out, but our existing skills and understanding of the game systems pollute our empathy. Some systems like multiplayer, economic or long term progressions yield surprising results or large play surfaces. Self-playtesting ends up being unreliable. So we need to rely on much less frequent cycles of playtesting with others.
Authored content exhibits varying degrees of leverage: Leverage is a measure of efficiency, how much value the content delivers relative to its cost.
Leverage = Meaningful contribution to the player experience / Sum of total authoring and tooling costs
High leverage: An evergreen piece of content (such as a National Anthem that that took hours to write and is used millions of times over hundreds of years) is high leverage.
Low leverage: A comic in a book that took days to draw, but is viewed only once and then forgotten is considered low leverage.
Other factors: The full cost/benefit structure includes the cost of set up the toolchain, the amount of content you make and the content pipeline everything needs to flow through. We’ll talk about these more in the Constraints section below.
Leverage is a useful concept used in planning, but understand that it is inexact. Once content hits an audience, they may choose to elevate what the author thought of as a minor element to evergreen status. There are scenes from a comic like Calvin and Hobbes that were just as expensive to create as any other scene, but their resonance with the audience turns them into a much greater experience.
Building content architectures involve an upfront cost: You need to pay for tooling. And learning the tools. And iterate on standards for your content. This is all before the team as author any shippable content chunks.
Traditional marginal media content costs are mostly linear. Once you’ve standardized on a chunk of writing, video or imagery, there are few meaningful economies of scale. The cost to create one comic panel is roughly the same as the cost to create a similar panel 100 pages later. Most efficiencies occur by descoping standard chunks and cleverly interweaving low cost chunks with high cost chunks.
In games, we can create non-linear content architectures: Content architectures can introduce non-linear leverage into the process of content creation. Such that for each additional hour of author labor, we get some more rich player experience out than if we had naively been making traditional content.
Diagram 1: Stages and costs of content chunk creation
This graph helps visualize trade offs.
A – Tooling Complete
B – Initial learning and prototyping cost paid, first content chunk created.
C – Break even on your fancy content pipeline. This is the first time all your work has a net benefit relative to just manually creating stuff from piecemeal.
D – Exhaustion sets in. Additional meaningful content is expensive because the player gains less value from each additional chunk of this type of experience.
Let’s say your goal is to create a high leverage content architecture. The first place to start is by understanding your constraints. I couch these primarily as questions a team needs to answer, since the answer will vary substantially based on the project. You’ll need to consider both sides of the leverage equation:
What is the cost of designing, building and testing the content?
What is the effectiveness of the content?
Cost – Prototyping: The goal here is answering the question “Would this imagined content deliver the experience we desire and how?”
What are your goals for this type of content?
How long will it take you to establish and explore the playspace limits for a particular class of content chunk?
What is the risk that this prototyping effort won’t pay off?
What are lower risk fallbacks if the prototyping fails to pay off?
Cost – Standardization: You need to create standards that eliminate edge cases and prevent the creation of weak content. This step is not free and often ignored.
How long will it take to create easy-to-communicate standards for the prototyped content?
How does the content fit into the content pipeline?
What tools are required to achieve desired efficiencies?
Cost – Iteration count on each chunk during production: Iteration is also not free and is commonly ignored.
How many implementation->playtesting->feedback iterations are required before the content chunk is polished and ready for the player?
Cost – Iteration speed: The speed of iteration typically determines how many you can fit. In my experience, the quality of content is directly correlated with the number and frequency of polishing iteration.
How long does it take to iterate on a chunk? Consider author iterations, where the author is testing based off their own playtesting perceptions. And also consider external iterations.
How can tooling be improved to speed up iteration?
Cost – Human resources: Each iteration heavy process needs to be designed, tested, optimized and mastered by living human beings operating at human-speed not computer-speed.
How many people across various disciplines does this chunk cost to make?
How long does it take people to master the creation of a chunk?
Cost – Technology: All that data only works because it hooks up into code.
What is the cost of the tech that supports the content?
Can you reuse or extend existing code when you add a new content use case?
What sort of dependencies and rigidities do certain tech choices create?
Cost – Game systems: Game play is a complex interlocking system of game mechanics and associated feedback loops. The content expresses and explores the playspace created by these systems.
What is the base cost of the game mechanics the content feeds?
How much content and of what types do game systems need to be fun?
How many game systems need to be in place before you can test the validity of the content?
How long does it take to balance the content across the various systems in order to test it?
Cost – Communication: As you add more people, their interdependence often increases the need to talk through design intent and issues. Hand-offs can be expensive or sources of blockage.
What are the hand-offs?
How do you make the hand-offs as efficient as possible? Where are blockage, delays or backlogs occurring?
Cost – Risk of failure: No creative undertaking as a certain outcome. Risk is converted directly into a cost in the form of rework or needing to implement an alternative design. For any specific class of content, you might not pay the cost, but over time the project as a whole will pay a higher cost for higher risk content.
What is higher risk content? Time and resources are two factors that have a giant impact. But the factor I’ve found most predictive is the past experience of the individual or the team. An experienced team will often know how much time and resources they need. An inexperienced team will be too busy exploring what they don’t know to budget effectively.
In order from lowest risk to highest risk
Content you’ve successfully made many times.
Content you’ve made 1 to 2 times.
Content similar to something you’ve made before.
Content that has clear playable examples in another game and you seek to copy the identical functionality.
Content inspired by something someone has made, but has not been demonstrated.
New content that has no direct analogue.
Note that there is both individual risk and team risk when talking about experience. If a task involves lots of people and they have not worked together before, they have a much higher risk of failure even if an individual contributor successfully worked on a similar project in the past.
One might think this sort of risk spectrum results in cookie cutter content. But that is not necessarily so, especially with smaller teams. A style of content produced by someone who has spent years working on an uncommon set of skills will often be lower risk than that same person trying to copy a popular style of content. Always consider the fit between creative skills and content, not just popularity or examples.
What is your team good at? What are they experienced at?
What content standards can you borrow from other projects?
What is the risk of failure for this chunk? Does it fit in a portfolio of risk?
Are your fallbacks if a prototype fails lower risk?
Cost – Late Revision: Only at the end of the project does the team start getting high volumes of quality player feedback. With live games, the bulk of the critical feedback will happen long after launch. So now you’ll need to update key load bearing chunks of content. What was ‘finished’ needs to be opened up, rebalanced, revised or completely redone.
Late revision is particularly problematic for games-as-a-service. If your initial launch is even slightly successful, the title will spend the majority of its life undergoing constant revision. The rigidity that you bake into content becomes a major constraint on the cost of future updates and whether or not your team can sustain the project. You live with it forever. Teams who only know single player games struggle here and need to reevaluate most of their assumptions.
What does it cost to change a chunk of content after it is finished and tied into all other dependencies? What does it cost to replace it?
What does it cost to change a set of content? What does it cost to replace it?
Diagram 2: Design insights happen throughout the schedule not just at the beginning.
Diagram 3: If your content pipeline is not amenable to late state changes, you’ll fail to capitalize on most of your design insights.
Total marginal chunk cost: So there are lots of costs that go into making a chunk of content. Be sure to honestly measure and summarize these. Blindly insisting on an optimistic fantasy helps no one.
After paying prototyping and standardization, what does it really take to call one additional chunk of content ‘finished’? Include iteration cost, human resource cost, communication cost.
In the cursed wail of every team edging like Zeno towards the finish line, what is the true cost of calling content “done”?
Effectiveness – Load bearing: We now can talk about the other side of the leverage equation. Let’s start with how some content is more important than other content. A game has pillars made of key experiences that it needs to deliver in order for it to be successful. This is the heavy weight of player, publisher and market expectations. Various mechanical systems and content support those pillars. Those that bear the most weight and would hurt the game most if they failed are considered “load bearing”.
It is also worth identifying content that is “non-load bearing”. These are places where you can use lower cost content. You might reuse existing content or apply generic purchased assets. Alternatively, you can use the fact that non-load bearing content is low risk in order to experiment and be playful. I often find some non-load bearing chunks like item descriptions and inject them with my quirkiest writing. Or give authoring of this content to someone who is learning. If this content fails, the game won’t fail.
What are the pillars of your game?
What content is critical to supporting those pillars?
Is a particular type of content load bearing? Or is it non-load bearing?
What is the fallback if this content doesn’t deliver on its promise?
Effectiveness – Optimal set size: No practical system is scale free. On one hand, you want this number to be as high as possible in order to maximize the prototyping investment. However, standardized content chunks also fade in effectiveness over time as well. There is often less marginal utility to a player as they experience the 200th level compared to the 1st level. And if you are crazy enough to make a 5000th level, the utility can turn negative. Some players start to see the patterns behind your standardization and will ignore or resent non-meaningful variation.
What is the size of the playspace this content addresses? Is it small? Is it large?
What is the sweet spot for set size where each chunk of content remains distinct and meaningful to the player?
Effectiveness – Resonance with real player motivation: This should fall out of the exercise of determining if content is loadbear, but it is worth treating as its own thing. The best content helps players fulfill their deepest intrinsic motivations. When content and system support support the various factors of self-determination theory, we see increased retention, engagement and player satisfaction.
Does the content facilitate competence? Does it help the player learn skills? Or feel a sense of growth?
Does the content facilitate autonomy? Does it help the player feel like they’ve chosen their path? Does it help them express their identity?
Does the content facilitate relatedness? Does the content connect the player with others who support them? Does it enable reciprocation loops that deepen relationships?
Basic Architectural Patterns
Now that you’ve got a bunch of knowledge about what type of content you need to make, you need to build the system that helps you make that content. Here are some techniques I think about when building high leverage content architectures.
Take these with a grain of salt. I find that as a team gains experience in a domain, they develop new tools and vocabulary custom tailored to the tasks at hand. So I encourage you to set strong constraints and then deliberately grow your team’s ability to experiment with and iterate on more efficient tools.
Each of the following tools will likely take your team a full game or two to start to understand and master.
Lego blocks: Embrace composition by building player facing experiences out of highly reusable standardized content chunks. Consider a non-lego block design like early graphical adventure games. Every pixel on the screen was hand placed. Every interactive puzzle was hand-scripted. Deep in the code there were common structures, but there was very little modularity or reuse.
Consider a game like Super Mario Bros. The world is composed out of standard block types, standard enemy types and standard player moves. Tiles are placed on a grid so their relationship to one another is highly predictable. The cost to create a screen of a Mario game is much less than the cost to create a screen of an adventure game. (Thankfully, no one measures gameplays by screen any longer!)
Modular blocks intended to be composed together are not limited to tiles. In the puzzle game Road Not Taken, each object was built out of a stack of standardized behaviors. A block might have the ability to be pushed. Or it might have another ability to slide if pushed. Or it could break. Or duplicate itself. Or move on its own. And by mixing and matching a relatively small number of these lego-like behaviors, we built out dozens of distinct objects.
What are the legos of your game?
What pieces of your games are not standardized building blocks? How might you turn them into reusable legos?
How do your legos snap together to build interesting compositions?
References: Lego blocks usually use referencing where this is a master object stored in some central location and then an instance of that content is used in the composition.
You may store instance specific properties. There’s a trade off here. In general you want to specify the minimum number of instanced properties as possible since global late revision that touch a 1000 instances are expensive. It is better to store the bulk of the behavior on the master so that if you can make a change in one central location, the change happens everywhere. However, some instanced properties let you adapt the instance to the current context.
What properties should be on the master?
What should be on the instances?
Templates: As you compose structures using your reusable chunks, you discover that there are some patterns you repeat again and again. Certain sub-elements might shift around, but there’s a recognizable boilerplate structure you keep needing to rebuild. To minimize work use templates, reusable structures that have blanks the author can fill in details.
Consider rooms in a Diablo-like game. There was a set of templates that defined each room. During level generation instances of the rooms would be plunked down and connected with hallways. However, inside each room a subset of different objects or enemies might appear. So even though there were standardized, reusable templates, each instance of the room felt different.
What are your common reusable patterns? Can you turn those into templates?
Which elements in those patterns can be varied in order to provide players with meaningfully different experiences?
Decoupling: As we’ve discussed, splitting content into chunks and assembling them into compositions creates dependencies. Dependencies aren’t always bad. References are a form of dependency where instances depend on the existence of their master. However there are many dependencies that increase both initial content creation cost and future iteration costs.
For example, recently we built a quest that required you to purchase an ingredient (onions!) from the store. The contents of the store were defined in chunks of data. While the quest asking for store items was defined in a totally different chunk. If the store didn’t have onions, the quest was not completable. Which just so happened to break the entire game.
This showcases some common issues with dependencies.
Difficult to spot: It wasn’t obvious looking at the quest that there was a dependency on the store. The quest config said nothing at all about where you get an onion and it was only by sorting through the entire config system we found the connection. I call this content pattern “Chunnel Design” after the famous tunnel that goes under the English channel. They dug the tunnel from both the French side and the English side with plans to meet up blinding in the middle. If either effort had been off, the tunnel wouldn’t have connected.
Expensive to fix: Instead of making a change in one location, we needed to make a change in multiple locations. With tangled dependencies, this can get quite expensive. In one project, we had to update 5 separate locations to get an item to show up in the store. A five tunnel chunnel. 🙂
Ambiguous ownership: The quest wasn’t able to specify anything about how a player gets the onion. And the store had no idea that someone might want the onion. Neither piece of content was responsible for making sure that the desired experience was delivered to the player. Even if we did fix the issue, it wasn’t clear we fixed it in the right spot. And the next time we fixed a similar issue, we might make a different decision. Which leads to edge cases and more unexpected problems later on.
Decoupling at the most basic level is the process of eliminating unnecessary problematic dependencies.
What dependencies are helping speed up authoring?
What dependencies are slowing down iteration?
Can you remove these costly dependencies?
Can you explicitly state dependencies in your data so they are obvious upon inspection?
Can you give ownership of the experience to fewer chunks, instead of spreading it across multiple chunks?
Can you add automated validation so you are instantly alerted when dependencies break?
Content pipelines: As you start to engage with both composition and decoupling, we start splitting complex content into stages of work. Early stages of work, composed of templates and referenced masters feed into later stages of composed instances. Each stage has its own required tools, processes for ingesting data from previous stages and processes for exporting data to subsequence stages. Put it all together and you’ve got a directed graph called a content pipeline.
Diagram 4: Sample content pipeline
A content pipeline might involve the following three sub-pipelines of character art, terrain art and behavior code feeding into a finished level. Notice that various content chunks pass through multiple stages in a fixed order across many tools in order to create the final output.
Directed pipelines have some interesting properties
Stages are composed in a fixed order: This ensures reproducibility of results. Selecting the right order is a big design choice that impacts your content production schedule. I often think of this as “up pipeline” and “down pipeline”. Changes at base stages cause ripple effect down pipeline. Changes down pipeline have fewer later stage dependencies, but have a linear cost to make change. Which can be a very expensive number if that surface area of content at the end of the pipeline is large.
Manual composition: Order matters so much because often the earliest pipeline stages are created and locked down. Then subsequent stages are built on top and the earlier stage is never changed. In platformers, designers build a chunk of player movement with locked jump distances. And then the layer of level design is built on top of this. Manual composition creates strong dependencies. Changing or replacing a locked stage invalidates the later stages. If those stages (such as hand crafted level) took time to build, naive changes to earlier stages can cause immense project thrash. Managing scheduling of locked stages is one big reason why we have producers and those miserable gantt charts. There are tricks to get around this issue, such as using stubbed in dummy data or placeholders. We’ll talk more about that below.
Automated composition. One way of reducing these dependencies is to automate the composition process. Procedural generation is one form of this. The rooms in a rogue-like are placed via an algorithm. If the rooms get bigger, that constraint is passed up to the next layer and the hallways connecting the rooms adapt accordingly. Unlike manual composition, the author can then make a change on almost any stage and the end content is rebuilt automatically. (Photoshop was so transformative because it pioneered automated layer composition in the visual arts)
Content at each stage can be referenced: Each layer is defined in a master chunk and instanced.
Automated composition + referenced chunks offers immense leverage by reducing the cost of authoring iteration. A content author can compose multiple layered compositions. And if late changes need to be made to ever base layers, it is less of an issue.
Observation – Non-linear leverage appears in how you build the pipeline: What we are seeing here is a key truth. Non-linear leverage in your content architecture rarely comes from how you structure your base chunks. Instead it appears in how you build the composition of those chunks. In my experience, the more you can move into hierarchies of composition, more leverage is available. This introduces its own complexity and cost so it isn’t a silver bullet.
Advanced Patterns – Manual composition
Sadly, it is rare that we can apply automated composition to every composition process in the pipeline. Anywhere there is manual composition, the order that elements are created matters. This presents some challenges:
How do you schedule work so the right stuff is complete before the next stage needs it?
How do you reduce the cost of making mistakes?
There are some common strategies. Any or all of these can be mixed and matched.
Vertical Slice: Build out a representative segment of the final content at full fidelity, test it to verify validity. Then meticulously lock down standards for each pipeline stage. In production, build content to these standards and trust that the end result will deliver on the promise of the vertical slice.
Issue – Slow iteration: However, building the vertical slice is expensive and leads to slow iterations. Imagine building out a whole level with complete mechanics and final art, discovering it doesn’t work and then throwing that away. I think of it as “Building the game five times” More often than not, teams get into the second or third iteration and are canceled.
Issue – Bureaucracy: Another issue with vertical slices is that it puts immense pressure on the standards. They must be perfect and they rarely are. The answer is often more documentation. This acts as an organizational tendency for large bureaucracies and large teams where waste is common. Due to rigidities in the system, change — when it does occur — is often a destructive coup or pogrom. Vertical slices are very common in AAA.
Bottoms up design: Identify most core “up pipeline” stages. Prototype them. Test them. Ensure they are fun. Polish them to a high degree of fidelity.
Now lock down that element of the design. Then move onto the next stage of the pipeline that builds on the locked down stage and repeat.
For example, if you are building a platformer, build, polish and lock down the most perfect jumping you can create. Then build a small level with blocks based off jumping so your game grows like an onion from the innermost layers. When you hear the advice “Focus on a fun core mechanic” it is usually a sign of bottoms up design.
Issue – Highly systemic games: An issue here is many games require multiple interlocking systems to be in place before you know the game is fun. Consider a game like Animal Crossing. It certainly has central mechanics like chopping trees and running around. But (having just worked on a game in this genre) until economy, narrative, pacing, affordances, inventory, other minigames are all in place, the game is desperately unfun.
Issue – Late stage changes: The other issue is again one of managing late stage changes. If you discover that you screwed up an aspect of the core gameplay early on, it can be expensive to pay the cost of that change rippling out across all the dependent layers of the content pipeline. An MMO (Age of Conan) baked the timing of their attacks into their female character animations. When community playtesting suggested they needed to speed these up, it was an expensive fix. The early assumptions baked into the content architecture bit them.
Placeholders: Build a vertical slice of your game, but fill it with low fidelity placeholder content. This lets you test the game quickly and identify issues. And since the placeholder content is relatively cheap to make, throwing it away doesn’t destroy your budget. As you become more confident of the validity of the work, you start refining and polishing.
Placeholders can be used with either vertical slices or bottoms up design and they inherit most of the same issues. Bottoms up design often results in piecemeal prototypes that don’t really tell you how the final game will play. Vertical slices still result in a lot of throw away work, but since you are using placeholders, iteration is much less expensive.
A version of the vertical slice + placeholder that I’m intrigued by is the “playable skeleton”. With this strategy, you create a full version of the game that is playable end-to-end as inexpensively as possible. And then you perform subsequent polishing passes until the game reaches a shippable state. Thimbleweed Park was built using a similar technique with a full playable version of all game rooms complete and iterated on before final art was added.
Issue – ignorant stakeholders: A common issue with placeholders is that stakeholders do not have the critical sophistication to understand what is placeholder and what is final. Games have been canceled when an executive looked at a graybox level and wondered why this game they are spending millions on is so obviously ugly. Many teams end up with a secret rule to only show their publishers near final art and claim it is placeholder. The risk of getting that one ignorant person is too high for honesty. And education can be an impossible lift.
Issue – weak player affordances and feedback: Players also don’t always understand placeholders. There’s an art to picking comprehensible placeholders that work well in a placetest; abstract boxes and colors are almost never the right answer. Instead go for lower fidelity content that is still thematically and symbolically representative. If you are supposed to be petting a dog, use a picture of a dog. You’ll learn important lessons iterating on the right affordances and feedback even in a prototype.
Scaffolding systems via value anchors: The challenge of cheaply validating systemic designs is unsolved. It is common, even when using vertical slices or playable skeletons to spend months (or years!) in the dark valley of faith as various systems slowly come online.
For example, in order to test a crafting system, you need to build the crafting system, a UI, add the crafting content, add sources for that content, balance the sources, balance the crafting costs and finally anchor the crafted items to a functional purpose within the broader game. Even if you build the base crafting functionality quickly, the other elements take a lot of time and effort to coalesce.
One approach is to stub in value anchors early in production. This is usually a large sink that’s easy to build but still gives purpose to the various content systems. By building the anchor first, you have something to judge the activities against. Later you can still add secondary activities and more nuanced anchors.
When you prototype an RPG, you can create a player level that is fed with XP. Then you can have various activities like combat feed into XP. Player levels feedback into power which in turn allows tackling of harder monsters. Later you can add additional skills, enemies and resources that expand the system. But you’ll always have something playable from early one.
Animal Crossing has a large sink in the form of paying bells to upgrade your house. Whatever activity you do results in items that can be sold to generate bells. This creates a simple skeleton to slowly add more activities, more resources and ultimately more player goals.
Anchors are a bit tricky to get right because they aren’t purely mechanical. They are about setting up systems of value and tie into deep player motivations. The reason upgrading the house in Animal Crossing is interesting is not because of the mechanics of upgrading! It is because the house holds your decoration and items, which in turn act as a signal of identity, progress and status. In our Animal Crossing-like Cozy Grove, we’ve built a prototype that had upgrading your ‘house’ without the decorating aspects. It didn’t anchor player value at all.
Advanced Patterns – Automated Composition
There are also content architectures that open up when you enable automated composition. This is an exciting open area ripe for additional experimentation and research. I expect over the next decade or two, we’ll see a steady adoption of content architectures with various forms of automated composition. Here are a few ideas that I’ve found helpful to get you started.
Thinking of procedural generation as an authoring helper: Broadly, many of our existing tools in this space are termed “procedural generation”. But this field has problematic roots.
Researchers and new proc gen developers look for magical algorithms that provide fountains of surprising new content. Like old cranks searching for perpetual motion machines, they hope to one day crack the problem of an infinite experience generator. It is very much the perspective of an engineer who is not an artist but still wants to magically create without learning art. Though certain machine-learning efforts show promise, I personally have no interest in this particularly philosophical approach.
How does it make the content creator more efficient?
Does your content author understand the tool?
How can they create richer content that resonates with players?
How can they reduce iteration time?
How can they decrease the pain of late changes?
It is these last two area where procedural generation techniques shine. A good automated composition pipeline allows designers to make changes at most stages and have those changes flow through into the end experience with little to no manual rework.
However, procedural generation has a very real upfront cost. You need to abstractly design about your content and how it is assembled. And build all the tooling for those specialized chunks. And then build the automation that assembles them. This can cost many multiples of just building a single content chunk manually. Long term, you accumulate benefits in terms of cheaper iterations, but it is rarely clear that the initial investment was worth it.
Technique – Combinatorics: Do you need 1000 chunks of content in a set? If so, the cost of making that content is often high. And the post-release cost of changing that set is likely high as well.
One technique is to split your desired content into sub-chunks that are arranged in orthogonal sets. And then use combinatorics to generate an expanded set of final content that covers a wider surface.
For example, in our game Cozy Grove, we have shells on the beach. This is split up as follows
Shell type: This is a small set of 6 basic types like clam, conch, whelk, starfish, cowrie, coral. Each of these chunks contains a set of properties for image, price, chance of spawning.
Shell season: This set contains 4 seasons and color variations across those seasons. It also contains filtering information so shells don’t spawn in the wrong season.
Shell rarity: A set of five rarities. Each contains modifications to chance of spawning and price. Additional information about which bitmap to use.
Master shell definition: This tells how these 3 orthogonal sets are to be combined. It also contains any properties shared across all shells, like behaviors or dusting value.
Once each of those is defined, there’s an automated composition step that combines them all together to generate 120 (6 * 4 * 5) expanded variants. This also provides us with non-linear leverage where adding one new shell type adds 20 new shells to collect.
Issue – Bowl of Oatmeal: Combinatorics make it trivial to create what Kate Compton calls Bowls of Oatmeal, vast amount of content that is neither perceptually unique or differentiated. Players will tend to latch onto patterns shared across your spread of content and filter out non-meaningful variation. The infinite yet weakly differentiated worlds of No Man’s Sky are one example.
There are a few techniques I’ve found useful here.
Choose smaller set sizes that don’t trigger player exhaustion. Small, highly differentiated sets are often much better than large undifferentiated sets. If you split your placespace up too finely, you get oatmeal.
Use cheaper content like names to obscure the rote nature of combinatorial expansion. One thing we do for shells is give every combination of season and type a unique name. That’s only 24 names and took very little time. And concatenating “rarity + 24 unique names” results in strings that feel unique.
Technique – Chocolate Chips Cookies: Another composition pattern is to mix high fidelity setpieces in a low cost substrate. You can think of your templated setpieces as chocolate chips. Players love them, but if they repeat them too often, they burnout on consuming them. So they must be used sparingly. And the substrate they are embedded in is the cookie dough. Pleasant, filling, endlessly edible. But not very unique or interesting.
Individually, these two types of content have flaws. The dough is low cost, but also results in bland experiences. The chocolate chips are high cost and overly consumable. But they provide great peak moments. By creating a pacing structure so that just as players are getting bored of the dough, they encounter a chip, the value of both can be extended.
In rogue-likes, you author setpieces in the form of rooms and boss encounters. And then you embed those in levels composed of randomly generated hallways and generic rooms. Just when you are getting tired of slogging through endless corridors, you see a magical unique room that changes the rest of your run.
Imagining the final experience, what aspects deserve to be meticulously authored? What aspects are filler?
What are your set pieces? Prototyping, standards, production processes and costs. How often can each one be used before players consume them?
What is your substrate?
What is the ratio of set pieces to substrate?
What is the pacing of setpieces?
Advanced Patterns – User content
There’s also a set of more volatile patterns that involve leveraging your players. You give up control and risk quality, but sometimes gain new sources of content far beyond the resources of your team.
Player sourced testing: If you have a strong pre-release community, you can ask them to test the game. This is perhaps obvious, but in the language of the model we’ve been discussing, it facilitates getting back rapid and rich feedback on your iterations. This path also includes analytics.
Player sourced game content: You can go further and source actual content chunks. The most common example of this is crowdsourced localization, but it can be extended to other types of content.
In Realm of the Mad God, we crowdsourced much of the pixel art. Some important lessons from this and crowdsourcing localization:
User friendly tools: Players don’t have the patience to learn typical developer tools.
Robust standards: You need explicit, heavily validated standards. Developers need a path creating the content right. Players need to be prevented from creating the content wrong. These sound similar, but the latter is a much harder requirement.
Credit: Acknowledge their contributions. This goes a long way towards encouraging them to help out. We held contests that were very effective.
Mods: Post launch you can open your game up to mods. It is quite common for long lived popular games to source entire expansion packs or members of the ongoing dev team from the mod community. It is a gift that keeps giving.
In-game social content: You can also build tools inside of your game and incentivize players to create content for other players. There are many variations of this, but the main thing to note is that good UGC systems require you to design your game around them. Not a simple add-on, but something at the heart of the core loop. Examples:
PvP: Players act as enemies for other players. Counterstrike, Chess.
Base builders: Players create bases for other players to destroy. Clash of Clans
Building games: Players cooperatively build in a space together. Minecraft, Factorio
Design games: Players create levels for other players to play. Super Mario Maker, Dreams
Meta: Designing tools
So far we’ve been mostly talking about how you design your data and the structure it lives within. But don’t forget that authoring this content is a human process; someone needs to create by hand the work feeding these magnificent pipelines. And for that you need great tools.
The goal of tools: Tools multiply the efforts of content authors. They help create:
Richer content: Tool unlock the ability to make types of content that were otherwise impossible or too time intensive to consider.
Cheaper content: Tools enable an author to create a chunk of content of a desired quality level more quickly.
More polished content: By reducing iteration time and improving feedback, an author is able to quickly polish their poor rough drafts into something that delivers
Unless you get into generative systems, they tend not to be used to create large quantities of new content from some base seeds. That’s more the role of combinatorics or other proc gen techniques.
All game tools are custom designed: The first and most critical lesson you should learn is there are no standard tools. Every tool needs to be custom tailored to best fit the following constraints
Skills of author: What level of abstraction does the author work best in? What affordances help them do their job? Game tools generally target intermediate and experts.
Requirements of the content chunks: What is the minimal set of data that should be hand authored to make an effective chunk?
Ingestion of the content: What is the efficient process by which authored content is connected up with the rest of the game?
Iteration requirements: How do the tools enable the author to make and see changes rapidly?
I suspect some of you are thinking, “But I have Photoshop! I have Maya! I have Unreal! Those are standard tools.” Sweet summer child.
Modern commercial tools are powerful enough to do almost anything. Without identifying and serving the previous constraints, you will flail. So like it or not, you still need to establish standard practices, procedures, naming conventions and automation scripts in order to use even something as ‘standard’ as Photoshop to efficiently build your specific game. There will always be a tool design process for each game, even if it is built on top of an existing tool chain.
A process for designing your tools
Constraints: Identify the four constraints for a particular type of content: Author Skills, Content requirements, Ingestion Pipeline, and Iteration requirements.
Initial Sample: Create an example of the type of content you are making. Get feedback from stakeholders if this is what you want to build.
Brainstorm building the sample: Talk to a real author. Not an imaginary one, but an actual person who is going to be creating these things. How would they build this? Is there anything that exists that could be leveraged? What are problems and workflows they imagine will come up? Small, cross functional strike teams are very effective if multiple people are involved.
Build a first version: Try for the 20% of features that gets you 80% of the functionality. Test the pipeline of creating and ingesting and seeing the content in the game end to end.
Get an author to use the first version as soon as possible: Have them make real content that is expected to be in the game. Listen to their complaints and dreams.
Fix issues: Fix as many easy issues immediately. Prioritize one or two big asks for the next rev. Repeat these last two steps until the tool converges on something ‘good enough’; it will never be perfect.
Mistake – Not basing the tool features off real content needs: The most common pitfall that plagues tool creation is that feedback and iteration steps (2, 3, 5 and 6) simply never happen. An engineer makes a tool. They (or antsy producers) declare the tool finished and the rest of the team is told to use it.
Often this first pass contains the wrong features.
Or weeks are wasted over engineering aspects that are unimportant.
Or they’ll have built in major workflow problems that are invisible to them because they don’t understand that X is an operation you need to do 300 times in an hour, not once per week.
In the best case, content authors don’t even use the tool and find cheaper workarounds that get the job done. You just lose the engineering effort. In the worst case, content authors use the tool but they spend truly enormous amounts of wasted time jumping through avoidable hoops. The result is typically bad, hacky content that was expensive to create. And often needs to be thrown away.
Mistake – Delays building real content: The next most common pitfall is that there is a large time gap between the first version being built and an author uses it to create real content. In addition to general problems of skipping iteration, waiting too long has the following negative effects.
Change becomes expensive. Code and processes petrify over time. When an engineer still has the code in their brain, feedback from the author is much easier to implement. Small tweaks happen quickly.
Authors are never taught how the tool works. An immediate dialogue between the creator of a tool and the content author inevitable results in knowledge transfer. So many times I’ve realized that there was a keyboard shortcut already implemented for a laborious task. But the conversation happened a month after the tool was built and the engineer had forgotten.
Tip – Shadowing: Content authors infest old tools like fungus in a moist fecund jungle. Strange content will seep out of every crevice in the toolchain. Wait long enough and you’ll see workarounds built off hacks forming the foundation huge swaths of your content. Authors learn, adapt and push tools in ways many find horrifying. In the process, inefficiencies creep in as the tool ends up being used in ways it was never intended.
This is normal. And it is a good thing. Clever content creators are discovering new opportunities and new requirements that couldn’t be predicted until they put a few hundred (or thousand) hours into actually building the desired content.
The first step in supporting your fungal creators is to understand how the tool is used in the real world. Shadowing is when a toolmaker watches a content creator build something. It is like playtesting for your tools.
Share a screen as a content creator builds something. If they start doing something strange, ask them why. The answers are delightful.
Record how long things take. Is anything surprising? A fun exercise is predicting how long you think things will take, and then compare it to reality.
Brainstorm ways of reducing iteration time. Can steps be removed or automated? Can automated steps be sped up? How would you make this process 10X faster?
Review standards: Do they need updating? Can edge cases or expensive exceptions be avoided going forward?
We’ve only managed to cover the most basic aspects of game content architectures. I hope you find enough here of interest to explore further. Observe your own projects with a critical eye, experiment when possible and share notes with others. For deep skills that cross multiple disciplines, a document alone will never be enough.
Be humble. Content architectures are not a magical silver bullet for making more meaningful content with less effort. They can be a huge pain in the ass that introduces immense complexity, costs and risk into your game. Because of the effort it takes to build and tune them, they often delay your ability to start playing the game.
Diagram 5: When each incremental chunk is expensive and you need a lot of them, a higher leverage content pipeline might be worth your time.
Learning curve: Any content architecture and toolset has a substantial learning cost. The specific team using the system needs to understand and practice building great content with the tools. I don’t mean to frighten anyone, but this can take years. A level designer who has been using Unreal for 3 years will generally be a lot more effective than one who has been using it for 6 months. A team that has been building content for a specific genre on a specific engine will be much the same.
Often the best tools and processes are the tools you know: I’m regularly amazed at how simple tools and simple content in the hands of experienced, talented teams results in world-class experiences. The content architecture of a novel isn’t complicated. Just a series of chapters composed of a few hundred pages. Authored using bog-standard text editors. Yet we give that to writers with years of experience under their belts and amazing work emerges.
A lot of times, you can just throw talent at a problem. And if you need to scale up, throw more bodies into the pipeline. This path is always an option as long as you’ve kept your content modular and highly decoupled.
The final constraint: Hand-authoring is our ultimate pinch point. Humans can only work as fast as humans work. They need to dream, experiment, clumsily and slowly make mistakes. It takes “human time” to have moments of insight and creative breakthroughs.
Naturally, as beancounters and producers, we want to multiply those efforts. To stretch out that costly thing and increase efficiency.
But this hand-authored content is also the soul of our games. Dilute it too much and you destroy the very thing that provides value. “More, Faster” is not better if you are churning out garbage.
Your content architecture is a delicate balance act. Where do you put all your limited, beautiful, messy, human effort in order to provide the highest quality experience for the player? A worthy design challeng e.
For Project Horseshoe 2019, an annual game designer think tank, our workgroup investigated how economics could help promote prosocial values. You can read the other reports here: https://www.projecthorseshoe.com/reports/
Attendees: Randy Farmer, Joshua Bayer, Tryggvi Hjaltason, Erin Hoffman-John, Daniel Cook, Ray Holmes
“What should young people do with their lives today? Many things, obviously. But the most daring thing is to create stable communities in which the terrible disease of loneliness can be cured.”
— Kurt Vonnegut, 1974
Multiplayer games can help build a player’s social support network. What would game design look like if our goals included reducing loneliness, decreasing toxicity and boosting a player’s positive connections with others? This paper looks at how we might use economics, an often dehumanizing and antisocial discipline, to support prosocial design goals.
What’s at stake
A multiplayer game can impact our player’s social health. By designing poorly we can do great harm. The two most likely negative outcomes are loneliness and toxicity.
Loneliness causes stress in humans broadly, relating to feelings of vulnerability, and can also provoke scarcity mindset, in which a host of negative outcomes occur. Scarcity mindset is a stress-induced “tunnel visioned” state that causes short term thinking associated with long term net negative outcomes.
Further amplifying the urgency of this problem is that increasing life expectancy is exacerbating loneliness. In a dark reinforcing loop, advancing age makes us more likely to be lonely, while it is known that loneliness poses particular health risks to the elderly. As the median age of the world population increases, so too will the seriousness of the loneliness epidemic. There’s an opportunity to be seized as ever increasing numbers of older people play games.
As is further explored in our appendix “Towards an action-based framework for mitigating loneliness”, we can rely on the heavily validated UCLA Loneliness Scale to provide a baseline measure for what we mean by loneliness (see more in appendix subsection “defining loneliness”).
It is a truism that people are mean to one another on the internet. There’s a growing recognition that toxicity in an online community stems in large part from weak social design combined with weak enforcement of positive social norms.
At the root of much toxicity is the misdirection of our human need to belong. When humans lack membership in healthy, eudaimonic organizations, they experience stress and seek to rapidly remedy the situation, often in long-term sub-optimal ways. They may fearfully lash out at others, imagining that putting others down helps them rise in status. They join tribal groups who use their shared pain to wreak havoc in the world in an attempt to control their feelings of fear and loneliness. Being a troll can fill an absence of purpose (as we will describe below, purpose is a core component of conquering loneliness) and this feels better to many than the isolation of not belonging. Toxicity is a rational (though naive and self-defeating) short-term strategy that emerges in the face a lack of human connection.
We often think of toxicity as bad people taking advantage of a poorly hardened design. (There is a small amount of truth to this theory; a tiny percentage of players are sociopaths.) As a result, we attempt to treat the symptoms of trolling and griefing with ever-increasing moderation or community management.
However, we are learning that a badly designed social system actively generates toxicity, often at a rate that will inevitably overwhelm the human resources aimed at controlling it. The systems can inadvertently isolate people in closed-off loops where their fundamental social needs are ignored. In toxic systems every new user is potentially rewarded if they adopt toxic behaviors.
Increasing social support networks as an overall solution
The broad solution to the bulk of both of these issues is to design systems that build relationships between players: preventing fire, rather than creating fire which must then be fought. If people are thriving, with strong social support networks, shared goals, and opportunities to grow, they’ll be less lonely. And they’ll be less likely to act out in toxic ways.
The grand project of prosocial game design
There are numerous pitfalls facing designers who seek to increase their players’ social capital. These are organized into at least three major categories.
Psychology: Humans require a series of well-documented steps in order to build friendships. You need the right sized groups of people, in correct density environments, engaged in mutually dependent reciprocal activities. If these psychological requirements are not met, players will always fail to form relationships. This topic is covered in detail in the 2016 Project Horseshoe report “Game design patterns that facilitate strangers becoming ‘friends’”.
Logistics: Human beings have rigid limits on how many relationships they can maintain and how long it takes them to form new ones. When we try to put players together into groups online, there arise numerous logistical challenges in satisfying all the spatial, temporal and psychological constraints. This topic is covered in detail in 2018 Project Horseshoe report “Design Practices for Human Scale Online Games”.
Economics: Online games are built upon an economic foundation of resources: their creation, transformation, trade and consumption. Almost all social systems interact with these economic systems. However, common economic practices (and their underlying theory) often unintentionally incentivize asocial or antisocial behavior. In particular, many of the key elements required by the psychological and logistical aspects of friendship formation are systematically undervalued within common economic practices.
This paper focuses on the final category: economic aspects of prosocial game design. We’ll cover the following topics:
The fundamental need for economic systems when designing multiplayer games;
The challenges inherent in applying economic theory to prosocial systems;
An approach to using economics by first defining prosocial values and goals;
Prosocial economic design patterns;
Dark patterns of economic design that sabotage prosocial systems.
Part 1: Economics & Games
Game designs always have an economy
When folks who have taken a course in micro or macro-economics think of economic design in a multiplayer game, they immediately imagine things like supply and demand, auction houses or player-to-player trade. And these are indeed classic economic systems. However, game design uses the term ‘economics’ in the broadest sense of the flow and transformation of resources, value, and incentives for player behavior.
First let’s discuss how game designers treat their economies, and then we’ll get into what we can learn from the study of economics.
A game designer’s definition of game economics
Almost all game systems that manipulate player incentives, acquisition of resources, or use of those resources, can be examined with an economic lens.
The practical version tends to take the shape of something similar to Joris Dorman’s definition of an ‘internal economy’.
The internal economy
When we build a game, we create a cartoon world that players agree to mostly operate within. Nothing within the cartoon world is real but we can still build meaningful relationships between virtual objects that give players an interesting system to manipulate.
The economic operations involving the creation and manipulation of endogenous value in the cartoon world are known as the internal economy.
Boundary between the real world and the game world
There’s limited permeability of the boundary between the real world and the cartoon world. You can think of this as the designer writing out the import / export laws for their bubble of play. For example:
A designer might specify that you add virtual resources to the game by spending real world currency.
Or they can specify a time-based limit to the rate at which progress tokens can be earned.
Or players might go outside the official rules of how to play the game and share tips and spoilers that they did not earn directly inside the game. They might find ways of transacting in real world currency for in-game goods against the developer’s policy. There is always a black market.
Elements of the internal economy
We have a variety of economic elements within a game. Many of these come to us via computer science and systems theory, not directly from traditional economics. They were subsequently adopted by game developers looking to name the nuts and bolts of game development they’d been refining for decades.
Tokens: Within this world, there are tokens that act as goods, products, or currencies. A token can measure or quantify almost anything, including abstract concepts such as attention, time or value. Properties of tokens can themselves become tokens (it’s tokens all the way down).
Sources: There are operations within the world that produce new tokens, sometimes connected to resource sinks.
Pools: There are pools that store quantities and types of tokens.
Sinks: There are operations, usually connected to locations or objects, that take tokens out of circulation.
Transforms: Some operations transform tokens in number or type.
Then there agents who operate mechanisms made out of these elements.
Players: Human agents who have various permissions for triggering different transforms, sources and sinks.
Black box: Computer agents or systems which also trigger various transforms, sources and sinks. Often the player’s job is to understand the systems of cause and effect hidden inside the black box.
With this relatively simplistic set of building blocks, a designer can model most economic flows within a game. This includes complex or emergent phenomena like various feedback loops, ownership (just another property of a token), or trade.
Economics drive game balancing
In additional to being the structural foundation that all systems design in a game rests upon, economics also impact player behavior via incentive structures.
We can ask questions such as: why does a player want a particular weapon in the game?
And then we can instrument and trace the flow of resources in the game.
And formulate a theory that the weapon has a utilitarian value in helping the player accomplish some other goal, such as killing a key boss.
And decide if the boss killing isn’t happening as frequently as we want, and so choose to make the weapon drop slightly more frequently.
This flow exemplifies classic metrics-based game balancing. And it is nearly impossible to do efficiently without spreadsheets and graphs tracking all the relationships of elements within the internal economy.
Economic systems are everywhere in games
Once you start looking, you’ll see economic systems everywhere. Consider the following common game system through the lens of game economics:
Leveling systems: XP points are tokens that emerge from the source of killing enemies and are in turn transformed into various skill tokens once they accumulate to a level cap in the leveling pool.
Items: A weapon is a token that enables the player agent to perform various transforms on the pool of enemy health tokens in order to generate XP tokens they own.
Chat: The chat channel has a budget of attentional and time resources it consumes with each text message token. Players generate the tokens and transform them into knowledge or relationships. If the pool of attentional resources is consumed by too many messages flooding in, there’s no attention left to process the messages and the information is lost.
Part 2: Challenges of applying economics to social systems
The thought that first came to mind when investigating the topic of ‘prosocial economics’ is that we should see what real-world economics has said about related ideas. Sadly, traditional real-world economics does not map perfectly to game economics. There’s been some impressive work exploring the overlap, but both the methods and goals of the two disciplines can be quite different.
Limited resources: There are limited resources in the world;
Unlimited needs: Greedy humans have essentially unlimited needs for those resources.
This leads to economists spending most of their time trying to answer a few big questions:
What goods and services do you (as a society) produce?
How do we produce them efficiently from our limited resources?
How do you deliver them?
Who gets them?
Right off the bat we can see there are some critical constraints that aren’t shared by game economics:
No scarcity: We usually don’t have limited resources unless we have deliberately decided to limit them. We can create as many virtual goods as desired whenever we want. Player attention and cash are the few limited resources we care about.
Virtual goods: Our goods and services operate in a cartoon world. We can imagine them to be whatever we desire — even infinitely transforming — as long as they fulfill their role in the internal economy or satisfy the player’s psychological needs. The movement of goods is not an issue unless we want it to be one.
Digital ownership: Anyone can own a given item if we allow them to. Just because Bob has the Magic Sword of Smiting doesn’t mean you can’t have one as well.
Economics only recently has embraced psychology and computation
Economics is an ever-evolving discipline, but it has theoretical foundations that reach back to at least the 1700s. The influence of older ideas and models continues to this day. For cross-disciplinary spelunkers, there’s simply a lot of economics history that needs to be parsed through in order to distinguish a modern, validated idea versus an ideological fossil.
Poor integration with psychology
Psychology wasn’t a thing, so early economic models of human behavior are problematic. Here’re just a few head scratchers.
Homo economicus: The most common behavioral model assumes that humans are atomic individuals who operate rationally and selfishly. We know now that humans have limited attention, are contextually altruistic, are highly tribal, and exhibit a wide range of irrational cognitive biases.
Individuals are the best judge of their needs: We know from more modern research that much of relationship formation and short-term valuation is obscured in order to foster long-term, mutually beneficial support relationships. Our brains are not conscious of the base psychological processes driving some of our most pressing needs and are thus unable to value relationships rationally.
Weak social modeling: Most economics models ignore basic human behavior such as friendship networks, affiliation networks, limits on cognitive resources (ex: attention) or altruism. Highly nuanced and layered group behavior is bundled up into cartoon-like institutional entities (The State).
Interesting areas of investigation include behavioral economics, which is starting to grapple with a few of these issues using piecemeal experiments.
Recent adoption of computational models and data collection
Modern economics study (since the 90s) has increasingly used computers to test more complex models. However, these build incrementally on early work that was limited by the data collection and computing capabilities of the time. Where game developers essentially have a panopticon that records every possible player interaction within our cartoon worlds, economists are usually desperately making do with any data at all.
Reliance on proxies: Often economics studies can track poor fidelity aggregate data (macroeconomic values) or smaller quantities of shallow data divorced from context (pricing or purchasing logs). They are forced to create proxies of scavenged proxies when attempting to describe the real world.
Weak sampling: It is expensive to sample everything in the real world so data is lost. Individual actions are often lost and there’s no way to get detailed or complete historical trails of underlying data.
In games, metrics and processing complex models is relatively cheap. We may gain more from studying game theory (especially iterated computational simulations) and some microeconomics. It is not currently obvious what macroeconomics offers beyond general rules of thumb.
The practice of economics erases many of the social phenomena we are interested in examining
Economics embraces reductive utilitarianism and posits you can put a price on anything. Once you make this critical assumption, there are all sorts of wonderful things you can do with prices, buying, selling, etc. However, simply putting public prices on relationship interactions breaks them.
Transactional relationships: We tend to be intrinsically motivated to connect to others and invest long term in a relationship where no extrinsic value is ever publicly admitted. When a relationship transitions into being an extrinsically rewarded transactional relationship, the relationship often suffers a catastrophic loss of value.
Undervaluing long standing social networks. Human groups create public goods in the form of unmeasured relationships, social norms and cultural practices. These public goods are incredibly valuable in terms of individual health and happiness. Yet they are not readily measured by economic transactions. Economics in general struggles with public goods, and social public goods are even more invisible.
Over-emphasis of short-term measurable improvements: Efficient production using measured inputs and outputs is an easily optimizable value. But it may or may not be a long term benefit. Unmeasured factors (aka externalities) often dominate social systems long-term. Politics, weather, technological change. And when these drive catastrophic failures, the response is usually “Whoops, well, we never promised we were perfect.” And tragically, the same tools with the same underlying flaws are redeployed for the next round.
A historical vs an experimental focus
The practice of economics is as much historical mapmaking as it is a science. Economists are mostly poking at existing, highly complex socio-economic systems and attempting to accurately measure results. Interventions intended to bring about future results are as much guesswork as they are predictions of proven models.
Reliance on natural experiments: Many large scale macroeconomic theories are essentially untestable in laboratory conditions. Instead, there’s a tendency to look for ‘natural experiments’ and fit the models to the data. Most important natural economic situations occur rarely (ex: major recessions) and are burdened by confounding variables. To a degree, models based on natural experiments become an exercise in data mining, with incentives for overfitting and with limited opportunity to invalidate resulting models. Some journals encourage pre-registering of study results in order to reduce p-hacking and other poor research practices, but this is not yet widespread or mandatory.
Underpowered experiments: Many economics experiments involve small groups (often college students in a class). Meta-analysis show that upwards of 50% of empirical economic studies results are not reproducible. And upwards of 80% overstate their effect size by 2-4 times.
Scope of experiments: Certain emergent elements of economies (such as firms or tribes) tend to show up only in large populations that run for extended periods of time. Social systems in particular are heavily mediated by group size and the relationship formation process that drives social effects collesces over 1000s of hours of real human interaction. Experiments done in a 45-minute classes or 3-hour labs are often poor analogues for long-term social economies.
In the last couple decades, there’s been an increased reliance on randomized trials (what game developers would consider a version of AB testing) and increased focus on confronting economics’ replication crisis. However, bringing scientific rigor to economics still appears to be a work in progress.
All this makes it a challenge to pull clear models out of economic literature and apply them directly to our game designs. At best, some microeconomic theories seem to be generally true within given contexts. But like many design tools, these are subtle instruments to be wielded to craft a desired outcome.
The practice of economics has increasingly become intertwined with the politics of governmental policy. Politics is as much a world of using the right rhetoric and building the right alliances, as it is about doing quality science with reproducible lessons.
Your typical theoretical economist will wait years, if not decades, before witnessing any policy changes that actively test their theories. And often the economists who are most successful are those that invest in the political relationships and rhetoric that makes their work palatable. This dynamic drives insidious corruption.
Scientist rhetoric: With the dramatic success of physics in unlocking the atom, economics was seen as a comparatively unreliable practice full of quacks, poor predictions and fake math. In response, modern economists cloaked themselves in complex, yet non-verifiable mathematical models and public claims of being truth speakers. “The math says X” acted as a powerful appeal to logic and credibility…whether or not the math was ever right.
Ideology: If economics is a design practice, what is the desired outcome being designed? More political creatures long ago realized that economics is a tool that can push a society towards a given set of social values. Various conservative and liberal political agendas regularly put forth economic policies intended to drive quite disparate futures.
Faux-economists: Much of the popular discussion of economics is delivered by political talking heads who have very little academic understanding of its limitations. They use scientist rhetoric and ideological propaganda techniques to push their agenda to the public. This problem is further exacerbated when political entities fund economic research.
The political influence alone makes it incredibly difficult for those outside of economics to distinguish if shared lessons are reliable insights or heavily biased propaganda. The latter has deep, deep roots that are often invisible and unquestioned to the more devoted practitioners of any given affinity group.
The conundrum of prosocial economics
So we are left with two conflicting thoughts after all our investigation
Economic design is required: We must carefully design, build and balance economic systems when creating multiplayer games. To paraphrase Douglas Martin, the alternative to good economic design is always bad economic design, not no design at all.
Economic tools are poor: The more academic study of economics gives us fewer reliable insights than might be hoped for. There’s a slight overlap between the crafting of economics of the real world and the crafting of economics of virtual worlds. But what insights might exist are heavily obscured by poor modeling of human behavior, weak experimental tools and a two-century deep cesspool of political propaganda.
We are not equipped to immediately solve this conundrum. There is clearly a vast project, far outside the scope of this paper, where those educated in the field of economics and game design dig through the dismal wastes of economic theory. Perhaps there are intellectual treasures to be found. For those future seekers, we’ve tried to document many of our unanswered questions in the Conclusion.
For the rest of this paper, we track back to our domain of expertise, game design. As game designers we can apply a prosocial aesthetic framework that helps us use economics ideas (if not the full economics discipline) to drive designed experiential results.
Part 3: Reframing economics as a tool for expressing a system of prosocial value
What if, instead of trying to treat economics as a science, we use it as a set of tools that support a design practice?
Game design aesthetics
A critical goal of game design is to create an aesthetic experience for the player. There’s some set of values the team is aiming towards creating.
Explicit values: Teams regularly write down their project pillar or an “X statement”. For example a team might decide to build a go-kart racing game that parents can play with their young children, with the hope that this will form warm family memories that will last a lifetime. There’s an audience, a desired outcome, and a set of values that the team works towards.
Implicit values: Games also are designed according to unstated values. For example, a 4X strategy developer might recreate the genocidal values of colonialism. When asked if this was their intent, they claim to have not thought about the issue at all; they were merely mimicking the practices of the established genre. However, unexamined ignorance of inherited values still results in a game with those values. Ignorance does not create ideologically neutral games.
Building, with intentionality, towards an aesthetic destination
With this design perspective, we have no need for rhetoric of manifest destiny or inevitability of scientist math. Instead, we are humbly upfront that designers will do the following:
Be explicit about their values.
Make design choices to the best of their ability…
…In order to create a player experience…
…That supports their selected set of moral values.
Now, the process is not entirely arbitrary. Since we are dealing with real humans containing real temporal, spatial, psychological and material constraints, this is an engineering exercise. We are craftspeople doing hard, practical labor. Game design can never be a purely theoretical fantasy or exercise in hopeful hermetic elegance.
With clear explicit values, it is possible to measure the result on our players and judge the success of our work. The machine that we build for our players either achieves our aesthetic goals or it does not. And we can then tweak and tune our system rules accordingly so that future iterations might hone more closely to our ideals.
Now let’s come back around to economics. What if we treat economics as a design practice instead of a science? One that is also seeking to create an aesthetic outcome? To express a set of designer-selected values?
In popular culture, this perspective on economics is uncommon. It is more likely to hear claims that there is One True Way of building an economy. Much of this is your basic rhetorical polarization, where if no one seems to be listening, you shout your opinion more loudly and with less nuance. Some of it may purely be a result of the relative youth of economics as a science. For now, however, let’s put the One True Way on the backburner.
Let us entertain the thought that there might be many valid economy designs, each of which deliver a particular set of aesthetic values. Our goal as economic designers is to craft the systems that drive our selected set of values.
Again, just as in game design, this craftsperson framing does not mean we are allowed to dream up any old fantasy. There are truths and common emergent dynamics in economic systems. Trade creates value! It also destroys it by missing key externalities. Supply and demand generally work! For certain types of goods and certain types of markets. There are selfish agents within any sort of exchange economy, as well as altruistic ones.
What sort of world do we wish to build and how does the economy we design serve those values?
We invite you to adopt an explicit set of prosocial values when you build your games and their supporting economies. These values include both experiences we want to build towards and outcomes we want to avoid.
Prosocial play involves players behaving in a manner that benefits the community as a whole. It is composed of many designed systems that facilitate the following:
Friendship: The formation and maintenance of healthy, meaningful friendship networks between players.
Thriving individuals: The facilitation of individuals’ eudaimonic happiness. Individuals feel competence, volition, and relatedness, both for themselves and for their friends.
Altruism: The promotion of activities that involve intrinsically motivated altruism and cooperation.
Positive group norms: The spread and enforcement of shared altruistic social norms within and across groups.
Shared goals: The definition and adoption of shared group goals. Players work towards those goals via mutual interdependence, and achieve feelings of purpose and meaningfulness.
There are also values we wish to avoid generating with our social systems.
Individual toxicity: Individual toxicity is when poorly socialized individuals resort to antisocial behaviors in order to meet their internal psychological needs. These are expressed as griefing, bullying, sociopathy, narcissistic abuse and other ego-centric patterns that put the individual above the group. These behaviors often provide short term benefits to the egotist and poor outcomes for everyone else in the community.
Group toxicity: Group toxicity is when intergroup friction (resource competition, enforcement of social norms, competing group identities) results in unhealthy interactions. While competition between groups can motivate group performance, it also can breed damaging behaviors such as organized aggression, hate crimes or racism. There’s a growing movement in games (centered around community management) dedicated to fighting toxicity of all types in games.
Loneliness: Loneliness is what a person feels when they want to be with other people, but cannot. This may be due to a sparse relationship network, or logistical challenges that prevent them from connecting. Loneliness is one of the key emotions we feel when our social support network fails. Like pain telling us the stove is hot and we should pull our hand away, loneliness tells us that our current social situation is untenable long term and we should seek out connection with others.
Benefits of prosocial design aesthetics
There are of course many potential values a designer might select. So why would we pick these specific prosocial values?
Support from psychology: There’s a growing body of research that suggests these values result in happy individuals and groups. This research (like all incomplete science) will no doubt grow and change over time, but it is the strongest experimentally verified foundation we’ve found for improving the lives of our players.
Ethics: Responsible social systems design requires an ethical core. Much like medical doctors, we are operating on human beings. Huge populations of humans, in fact. As ethical practitioners, we need our own equivalent of the Hippocratic oath. We should do no harm, and if possible, improve the social health of our players. The prosocial values listed are an attempt at creating a code of conduct that is a minimum ethical bar.
Part 4: Prosocial economic design patterns
The following is an incomplete set of economic design patterns. Like all design patterns, they provide the canny designer with early tools for supporting prosocial values in their game using economic systems. Be careful; patterns do not guarantee results. They are instruments to be wielded with skill, precision and craft. If you want to get the result you desire, each of these patterns benefits from a lifetime of intentional practice.
This set of patterns is by no means complete. There’s immense work to be done exploring and extending these ideas through hands-on practice and iteration with live game populations. But it helps to begin somewhere.
Pattern: Friendship formula
To start with, we need a richer psychological foundation to build upon than economists’ rational optimizer. A useful model for social systems design is the distinct process by which human relationships form. This contains elements of contextual reciprocity that are found in across multiple field: Social psychology, newer flavors of both economic game theory, economic altruism models (see Appendix) and behavioral economics. We can use the friendship formula as a key tool to design prosocial values into a game.
Key friendship factors
Though every friendship has a unique history, they all require several key factors
Proximity: People must be able to interact with one another.
Repeat encounters: The same people need to identify and interact with one another repeatedly. For example, matchmaking systems that match different strangers together are weak social systems because there will never be enough repeat encounters to facilitate true friendship formation.
Reciprocity: People exchange resources with one another in the form of attention, conversation, gifts and support. One party makes an overture and the other party responds. This is an opt-in process at every stage. Successful reciprocity loops tend to increase in value and effort over time.
Disclosure: As the relationship grows, people will disclose intimate information to one another. This helps build trust and fine tune mutually negotiated social norms.
The accumulation of trust
As the reciprocation process continues, the participants gain trust in one another. The higher the trust, the greater the strength of the relationship. Ultimately highly trusted friends form key long-term support networks in times of need. It is a blind investment for most people; with each short-term interaction, they don’t fully know why they are investing. Yet long-term the strong support network predicts improve health and a longer, more satisfying life.
Accelerator of friendship formation
There are additional factors that help friendships grow more quickly.
Similarity: If two people feel they are similar to one another, they are more likely to initiate the friendship process.
Intensity: If two people are in an intensely emotional situation due to high risk, large amounts of resources in play or time pressure, they are more likely to become friends.
Autonomy support: Self-determination theory suggest that people who support one another’s autonomy needs (the need to feel like you are choosing your path in life) are more likely to invest further in the friendship.
Pattern: Measuring trust
A key metric of relationship strength is trust between two individuals. As trust in a community increases, support networks flourish. However, trust is typically considered an externality, a factor poorly measured or valued by economic systems. So it gets optimized out in the name of efficiency. By measuring trust, we help ensure that it is treated as a first class citizen alongside more material concerns.
What to measure
Trust is an internal factor that cannot be measured directly so instead we need to rely on proxies. These won’t be 100% reliable (and represent a major area for further discovery and research), but are a starting point.
First, measure pairwise bonds: We’ll start by looking at player dyads. This is the most fundamental network connection and more complex network topologies can be derived from this data. There are two different (asymmetric) bonds for any given pair of player; the bond player A to player B and the bond from player B to player A. For simplicity, many proxies collapse this into a single symmetric bond.
Active time spent together: One of the easiest symmetric bond proxies is simply tracking if players are in the same area together. Due to players being idle, it is usually wise to track time only if they are actively playing. This metric does not track the intensity of the trust, but merely the duration. Higher values of time spent together means your dyad fulfills the initial requirements of the friendship formula (repeat, serendipitous encounters).
Success together in high trust situations: If your game has cooperative or team activities, you can track if various dyads are successful and how often.
Time spent talking positively: In games with text chat, we can go one step further and track number of times that a player talks to another player using positive language. Tracking simple words lists with positive affect are a good baseline and there is more sophisticated language analysis available if necessary. There are versions of this technique that can be applied to vocal analysis as well to determine positive or negative tone. The nice thing about this metric is that it tracks asymmetric relationship bonds, where one person talks a lot or uses relatively more positive language than another. Time spent talking, especially on private channels is a sign of increased intimacy and is almost always correlates with high trust.
Analysis and tuning
Once you have these metrics, the development team uses them to tune the system. This is for the most part standard data analysis; you create baselines and then track to see if any of your design changes are impacting the baselines positively. Note that these metrics are not usually player facing for the reasons listed below in Challenges.
Total trust over time for each dyad. Trust should accumulate slowly over hundreds or thousands of hours. Total trust is a reasonable proxy for overall strength of the pair’s relationship.
Rate of trust accumulation. Large gaps in the accumulation of trust suggest a pause in the relationship. Humans have a limited budget of active relationships so it is very natural for someone to stop reciprocating with another player as the relationship fades. Gaps let us track the length of the relationship. As well as if the relationship is rekindled at some later point.
Create trust metrics for higher order groups: Total trust can be aggregated from dyads and tracked for an entire guild. Or a cohort that comes in from a particular onboarding source. You can often correlate these with other variables like guild churn.
In Steambirds Alliance, a cooperative MMO, we measure trust by a ‘togetherness’ factor. When a player kills an enemy, all nearby players also get XP for the kill (a positive sum resource as in Pattern: Positive Sum Resources below.) This event is tracked and stored on each player as a list of other players nearby that also got XP. We do initial tracking on the client and then send periodic lists to our metrics server. We post-process this data to generate various graphs.
So how would we categorize the strengths and limitations of this metric?
Symmetric bond: Due to how XP is given we don’t know if one players has higher trust than the other. So this metric is limited to assuming that both players like one another the same amount.
Mixed trust: We’ve got a weak tracking of “success together in high trust situations.” There’s some nuance. Players who are getting XP from fighting smaller enemies need lower coordination so trust doesn’t need to be as high. But players who are getting XP from higher coordination boss battles are likely in a higher trust relationship. In the future we could split those two metrics and test if the increase in fidelity helps us better track useful behavior.
Actively playing: We assume that when players get XP, they are actively playing (since inactive players don’t get XP) and they are working with others to blow up their mutual NPC enemies.
No accounting for freeloaders: We don’t account for free-loaders or bots because we haven’t observed them being a meaningful population of players.
A mistake we made early on was looking primarily at metrics like retention and monetization. These simply don’t tell us much about what motivates players to play. If I were to build the game again, I would have implemented the togetherness metric for the very first private alpha. Player relationships are top-level intrinsic motivators and by only measuring them late in the process, we completely misunderstood the state of the game we were building.
Challenge: Avoid sharing pairwise trust metrics
Imagine if your friends all had a trust score hanging over their head. And when you do some small things, you witness the number change. Your relationships would suddenly become transactional in nature with clear extrinsic motivators in the form of your willingness to make that number go up or down. And we know transaction relationships and extrinsic motivators reduce trust. Are your interacting with your friends because you like them and they like you? Or are you trying to make a stupid number move?
So never share detailed trust metrics with your players. Trust, like many social variables, if revealed to the observed subject as an operationalized metric irrevocably changes the subject’s behavior. You’ll ruin the validity of your metrics and likely degrade the relationships between your players. (This is also one of the reasons why ‘likes’ in social media end up being a source of toxicity and in general a very poor practice.)
Challenge: Sharing group health
You can, if needed, share some high order group health information. The best practice here is to keep it vague, heavily delayed and multi-dimensional so that the underlying metrics cannot be easily gamed. A common use for group health information is to drive positive behavior by directing players towards a few key activities that developers know will improve overall social capital. Think of directives that are broad like the Ten Commandments so that players maintain agency and localized judgement. Avoid suggesting highly specific (and thus identifiable and gameable) activities.
Challenge: Trust differs across social contexts
Individual trust exists on top of a bedrock of group norms. For example, a pickup basketball team is a high coordination, moderate trust group. Players know that within the context of the basketball court they can trust one another to play according to the social norms (the rules) of pickup basketball. If you only sampled this social context, you might imagine that everyone playing is in a deep relationship with one another.
Yet, this relationship is contextual. Outside the basketball court, two players might never talk. When you create your proxies for trust and social capital, it is worth taking into account context. The more rigid and proscribed the rules of group coordination, the less actual trust is required for players to work together. And your metrics of trust may not travel to other portions of your game.
A way around this is to track multiple trust metrics in multiple contexts. High trust dyads in multiple context should be treated as having stronger relationships than those with high trust metrics in only a single context. Note that one of the more interesting to measure contexts is family bonds or mate bonds. These are often have large impacts on behavior but are rarely perfectly visible from inside the game.
Pattern: Positive sum resources
Positive sum (also called non-zero sum) resources are a key economic tool for ensuring cooperative play.
Zero sum resources
Material resources in the real-world are zero sum resources. If I own a piece of pie, there is one less piece of pie for you to own. If I consume that piece of pie, it is lost to you forever. This probably makes you irritable due to loss aversion. An economy of zero sum resources is a world of scarcity. The challenge economics attempts to solve is how we might split up these limited resources in an efficient fashion. Inevitably this involves some form of competition either via trade, negotiation or warfare. All of these tend to reward (at least in the short-term) selfish strategies and their resulting social toxicity.
Positive sum resources
However, in digital worlds, resources are mere bits. Making more of a resource is free. If we found a positive sum digital pie, you could have a slice and I could have a slice and the pie would be undiminished. My getting a piece does not prevent you from getting a piece. There is no need for competition between two parties over a scarce resource. This area of exploration is connected to the software theory of agalmics: non-scarce resources.
There are a few natural positive sum resources, and correspondingly game systems based on non-scarce resources are — scarce. Time, for example, is something that everyone experiences equally and simultaneously (it is also not transferable). Information is usually positive sum. If I read a book, you can too.
With code, we can make almost any resource positive sum. When a monster drops loot for one player, it can also drop loot for any other player that did damage. Whether or not players compete over a resources becomes a design choice, not a fundamental constraint.
When doing prosocial designs, positive sum resources are one of the first tools you should reach for.
Watch people play and look for moments of competition or toxicity.
Are there zero-sum resources at the heart of those interactions?
Can you turn those into positive sum resources in a way that players are no longer incentivized to be toxic?
Challenge: Building games around positive sum resources
If you are new to game design, you might imagine that games require zero-sum competition or at least a sense of winning to be enjoyable. Luckily there are many classes of gameplay that work with positive sum resources
Coordination and cooperation activities: Players with differentiated and limited capabilities can work together to accomplish goals larger than themselves.
Races: You can still have competitive challenges where players try to do their best within some limited amount of time or a fixed number of moves. Most cumulative scores are an inherently positive sum resource where anyone can earn points independent of other players.
In general, almost any Player vs Environment (PvE) game is amenable to being redesigned using positive sum resources.
Challenge: Infinite sources and imbalanced economies
If everyone gets resources, how can we prevent our sources from generating too many resources and flooding the world with abundance? We often rely on scarcity to creating prestige tokens or tune the pacing of gameplay.
Per player caps: Capping the number of harvestable positive sum resources per player restricts the flow into the world. A tree might be harvestable for apples, but any single player can only harvest the tree once.
Per group caps: We can also cap total number of harvestable items per group. This helps prevent intergroup competition by ensuring each group has equitable resources.
Transaction costs: Even with caps, the total number of items increases for each additional resource receiving entity (individual or group) in the world. With cheap trade transactions between entities, it is possible to pool vast numbers of resources. A thousand players means a thousand times the items. However, you can prevent global pooling with large transaction taxes. Transport might be expensive or you could charge a hard currency as a trade fee. This ensures that players are encouraged to use the resources locally vs globally.
Appropriate sinks: Any flow of resources, no matter how large can be balanced with large enough sinks. Measure the rate that positive sum resources are coming into the world. In real world economics this can be a tricky thing to figure out. In a virtual world it is a simple metrics check. Then tune sinks such that an equivalent is sucked out. For example, you might find that 20,000 new Magic Coconuts are flooding into the economy each day. A decay rate of 24 hours ensures that this (newly) highly perishable good never accumulates more than 20,000 units.
It is important to internalize that as a game economy designer, you control the sources, the sinks and the narrative justification for why the world works as it does. Scarcity as well as abundance are aesthetic choices.
Challenge: A human’s total relationship budget is a zero-sum resource
We might imagine that relationships are also positive sum resources. Me being friends with you shouldn’t have any impact on whether or not I can be friends with someone else. The reality is complicated.
In a highly local context, when you consider a few people at a time, forming a new relationship creates a positive sum public good. This is shared between the people in the relationship and essentially creates value in the form of social support and improved coordination. It is often beneficial to make overtures to weakly connected players you encounter on a regular basis.
However if you zoom out and consider the entire social network of an individual, they have limited social resources to spare. The social psychology concept of Dunbar’s Layers suggests that humans have a relatively strict budget on both the total number of relationships and the number of high strength relationships their brain can manage.
For someone with a full set of friends, investing in relationships in one layer pulls resources from from other layers.
Like many social resources, this is a difficult-to-acknowledge trade off. By explicitly acting upon that ideas that total social energy is zero-sum, especially in localized small group settings, the relationship become codified and transactional in nature. And thus suffers a drop in trust.
Pattern: Knowledge Resources
Nobel Prize winner Paul Romer has looked at a specific form of positive sum resource known as a knowledge resource. By taking a particular set of scarce zero-sum resources and performing learned transformations on them, we can derive vastly more utility than if we had just used them naively. For example, wood might be burned in an open fire pit to create the desired resource ‘heat’. However, if someone knows how to build a brick stove, we can burn the wood hotter, store heat in the stone and ultimately gain more heat from less wood. From this perspective, knowledge is positive sum resources that help dramatically increase the efficiency of using scarce goods.
A wonderful prosocial attribute of knowledge goods is that supply is determined by the number of clever people you have creating them. Since knowledge is research by clever people, the more clever people we have playing, the more knowledge we’ll likely gain.
This is the opposite of most zero-sum scenarios, where having more people around drives increased competition for scarce goods. With appropriate design of your knowledge good economy you can make it so instead more smart players are an advantage, not a threat.
Some examples of knowledge goods
Player skill: When players learn actual skills such as learning to execute a tricky fighting combo or defeat a complicated boss, they’ve acquired a knowledge good. In terms of interaction loops, this involves combining base level interaction (like jump in Super Mario Brothers) into compound interactions (such a using a double jump to get past a tricky section). Player skills, like most knowledge goods are teachable and create an economy of attention and narrative around passing skills efficiently onto other learners. Streamers traffic, in part, by showcasing their knowledge goods (they also foster parasocial relationships and acts as reference relationships, but that’s a whole other set of topics)
Technological process: Conceptually similar to player skills, technological processes involve combine existing resources and tools using the right order, amount and methods in order to create more useful resources and tools.
Explorable spaces: If you have maps full of unknown areas or hidden secrets, the act of exploring the space yields knowledge. You can share this information with trusted allies.
Spies and espionage: Games like Eve set up competitive games, where knowing where and when an attack or resource transfer occurs makes the difference between success or failure. This adds a layer of trust and betrayal where sharing knowledge goods with the wrong people may result in your downfall.
Challenge: What about virtual knowledge?
Video games have a long history of creating tokens that represent real knowledge. Instead of actually training to gain the skill of fighting with a sword, instead players are awarded with a virtual token (or virtual skill in Jesse Schell’s terminology) that says they can now fight with a sword. Or at least perform the themed in-game action that looks like sword fighting.
Virtual knowledge is a straightforward game resource that we can choose to make positive sum or not. Since it is just a token, our systems can trivially pass it around or give it to various players. As unlocks, items or whatever.
To make virtual knowledge more social, you need to build in some form of transfer mechanism between players. Real knowledge goods have an implicit transfer mechanism in the form of conversation, but that doesn’t work for virtual knowledge. In MUD of yore, the only way to learn a game skill was to approach a skilled player and ask them to ‘trained’ you. Usually for a fee or time. There were fun variations where an advanced player can only advance further if they manage to teach a newer player one of these virtual skills. These creates a tit-for-tat reciprocation loop where both players are getting something they need.
In general, when setting up transfer mechanisms, such as the one here with virtual skills, try to create a natural interdependency between players. Economic mechanisms that encourage players to seek out and interact with other players helps facilitate the friendship equation.
Pattern: Voting Resources
A specific form of positive sum resource is a vote. Each voter has an explicit ownership of their vote and there are usually rules to prevent vote selling. If more voters appear, using the magic of positive sum resources, they also get a vote.
Votes are then transformed via a decision mechanism (aka voting) that determines whether or not some course of action is taken. Voting is a social system for managing politics.
The important thing to note here is that we typically don’t think of votes as economic resources. They are often talked about as part of the domain of political science and most literature covers a handful of relatively conservative systems (plurality, ranked voting, etc). But once we reframe them in economic terms, we gain a large number of tools for manipulating and building novel prosocial voting economies.
In the multiplayer VR game Beartopia, players could build various communal projects for their shared virtual village. However, there was a limited amount of public space and it was undesirable let anyone simply build what they wanted without buy-in from other players.
So we designed an obfuscated voting system themed as crafting.
If a player did a minimum amount of work, they could harvest berries from a bush. Berries were a crafting ingredient used in various communal projects.
Secretly behind the scenes, the crafting ingredients were considered to be positive sum voting resources. They were instanced per player. They were untradeable. They had a cap on how many could be accumulated.
When a communal building project came up, players needed to pool their crafting resources to complete it. The was the equivalent of needing to vote for the project.
The cost for large projects was determined by the number of players in the world. It was tuned so that it would require a substantial amount of participation by players the world to create long-lived objects that consumed public space.
We also had short-term public objects that almost any individual could create. They might take a day of individual harvesting labor but also expire in 8 hours. Since these were not a lasting consumption of the public good (communal space), we could set the crafting thresholds lower. Thus allowing more individual control on short time spans. We wanted a system where individuals were empowered to make local changes, but you needed the political power of increasingly larger groups to make more permanent communal changes.
By putting expiration dates on most public projects, we ensured that with a lack of ongoing public attention, public goods would revert back to the public domain. This helped create persistent long term shared goals for players simply seeking to maintain the status quo.
Pattern: Interdependency of player roles
One of the early lessons of the industrial revolution was that division of labor allowed workers to vastly increase their productivity at multi-step tasks. Groups of specialized workers working together were more productive than an equally sized group of generalists.
This pattern for organizing human resources has three interesting attributes that make it pervasive throughout social systems design.
Increased efficiency: Functional interdependence is relied on economic incentives; the specialized players are more effective or efficient a certain types of resource acquisition or transformation. In other words there are natural incentives readily obvious to the individual player that they should specialize in order to maximize their personal playtime.
Coordination required: Interdependency leverages some form of process or skill (a knowledge good!) to coordinate all the players together. To gain the benefits of specialization players need to interact with others. And teach one another.
Trust-based: Finally, it is a social system fueled by trust. Each member of the group needs to have trust that other members of the group will perform their roles according to the plan. There are downsides for the individual members if the group falls apart. Specialization has a substantial opportunity cost; such players are unlikely to be effective generalists. In the real world, if the societal coordination systems that let specialists (most game designers, apart from Jason Rohrer) work together failed, we would all starve.
Challenge: High performance, specialized group activities scares new players
Because there are strong penalties associated with the failure of high coordination activities, it takes a huge amount of group trust to pull off the most efficient (and complex) activities. One of the biggest fears of new players is that they’ll be required to engage in specialized, high coordination group performances. If they fail, their reputation with this new group of people is tarnished forever. Leading with high risk, high coordination activities generally will send folks running from your game.
So designers need to build a ladder of activities in their game, starting with low trust activities between generalists and moving towards higher trust activities between specialists. The following illustration from the paper The Trust Spectrum shows the basic progression.
One response from systems designers is to build systems that allow coordination between specialized players at lower levels of trust. The thought goes, “Since trust is rare and hard to acquire, perhaps we could get efficiency out of our specialized groups in more mechanistic and scalable ways”
In the real-world, we’ve seen this in a practice known as deskilling, where highly trained skills are turned into a series of rote actions that are simple to perform and teach. These deskilled actions are coordinated, not by trust, but by an algorithmic (often computerized) system. A very early version of this was the assembly line. These systems scale to larger groups and can make use of broader labor pools. If you only care about the output of the system, they can be quite attractive.
However, if you care about the experience of the players, there are a couple questionable things happening here.
Deskilling removes the opportunity cost of specialization. There is no investment in a given role. Role switching cheap.
There’s less trust required to coordinate. The simplicity of the actions plus the role of the computerized coordinator means that groups practicing together see less overall efficiency gain.
Each worker becomes a low-trust cog that is easily replaced if they do not do their specialized role.
Deskilling systems are typical low trust systems that are helpful to new players. However, they are unable to facilitate the formation of high trust relationships.
Pattern: Shared Vulnerability
We know from psychology studies that shared pain acts as “social glue”: experiences of shared struggle create tight bonds of trust that yield greater social cohesion and measurably improved cooperation.
This can be harnessed in games through structuring experiences whereby players experience shared struggle early in the formation of a group. In many online games, players have discovered this organically and include it in their guild rituals.
Example: Guild Onboarding in Eve Online
When high-retention fleets were studied in Eve Online, a pattern emerged in the fleet manuals (often exceeding 80 pages) created by these high-functioning organizations.
One particularly high functioning fleet created a formula that was to be followed exactly:
Recruiter finds new recruits in a neutral zone
Recruiter gives everyone the same ship
Recruiter gives each new recruit a role that they understand. The intention is to make sure the recruit feels useful to the larger whole.
Specific behavioral instructions are given, for example: “A. Find this target, shoot it.” Then “B. Shoot targets of type N that approach.”
Some organizations specify redundancy in roles, so that there is no identifiable single point of failure in the mission.
Then the newly formed group is taken into combat with the specific intention that they will all die together.
After this happens, the recruiter explains that it was okay, and that it was a bonding exercise.
The recruiter also replaces all lost materials. The intention is to demonstrate support in a time of need.
This playbook of an experience creates high retention in groups through this mechanism of a shared memory and an establishment of interdependence, loyalty, and generosity.
Challenge: Formalizing Trauma
The risk of relying too heavily on this is that it creates downstream undercurrents that influence a game’s overall culture by grounding the bonding event in shared trauma. If these traumas are significant enough, they can convey lasting damage onto the social relationships of the group members. We can assume that most in-game traumas are far less significant than real world traumas, but these experiences fall into an unstudied place and it can become hard to determine how much pain is too much.
It seems possible to assume that the infliction of fear and experience of loss will have some repercussions on the group’s future dynamics. Further, the coping mechanisms that develop under crisis may not transfer to more peaceful contexts. Therefore the shared trauma of an early experience may have to be continued thematically through the game (a game about war continues being about war) which then sets a dynamic across the experience that is hard to disrupt. The challenge then is to carefully calibrate the kind of shared vulnerability — which is likely a very wide design space — and manage a thoughtful transition to more peaceful forms of gameplay that amount to recovery therapy.
Challenge: Solidifying Out-Group Hostility
Because these mechanisms for bonding are explicitly successful within guild contexts, which are tribal contexts, it is not clear whether the benefits persist or are possible without a kind of enemy tribe. For combat-based games, this is highly effective, but it isn’t as clear how it would translate to a non-zero-sum shared massive environment. Out-group/inter-tribal hostility is a powerful design mechanism in and of itself that is of questionable prosociality — being in a kind of “adrenaline” category of design mechanism that results in very high levels of comfort within established groups and higher isolation and discomfort outside of or between groups.
Pattern: Player-to-player Trade
Trade increases overall value by allowing exchange between players who own differentiated goods. By giving up something a player doesn’t need for something that players does need, both players in a transaction come out ahead. There’s a lot to say on this topic; many mistake this topic for the totality of economics. For a very brief overview, see the appendix on Trade.
From a prosocial perspective, the question we are interested in is “How does trade improve human relationships?”
On a macro level, by creating abundance through trade, we theoretically can escape a world where poverty-level survival dominates our daily lives. Excess resources could then be invested in our relationships. The not dying from famine, war, disease, ignorance and other outcomes of extreme scarcity is nice as well.
On a micro level, trade results in the reciprocal negotiation between at least two traders. As a result of this negotiation, people inevitably disclose their values with one another. And perhaps start to form a relationship. This style of trade is most common with ‘less efficient’, person-to-person barter systems. Though claims of efficiency depend on large part on what is being measured.
Challenge: Auctions dehumanize buyers and sellers
One of the great inventions of modern capitalism is the ability to boil down all of a person’s values into a single price on a commoditized good. A buyer can decide if they are willing to pay the price and the seller (by listing the price) automatically agrees to the subsequent sale.
Auction houses turn both the buyer and seller into low-trust, mechanistic entities. They can engage with a regularly updated listing of goods, quantities and prices and ignore the human on the other side. Humanity, in the form of face-to-face reciprocal interactions between people with names, histories, desires and culture, has been meticulously eliminated from the process. Too inefficient.
This results in immense improvements in material market efficiency. Selfish players clamor for such features in any game that includes trade. But it is worth asking if it drives the prosocial results that we desire.
Before you add a global auction house, consider the following ideas:
Games don’t require material efficiency. We have enough control over our economies that we can generate abundance on demand if required. Consider, players regularly ask for infinite power or inventory and we don’t give them those things. Because we know scarcity is a design element that drives our experiential outcomes. Similarly, there’s no law of nature stating we must climb an inevitable ladder towards ever greater efficiency in order to satisfy infinitely selfish actors.
Build barter or gifting-based trade system for friends: Barter systems are disliked because they put pressure on our limited trust and relationship budget. Negotiating prices with complete strangers can be time intensive and exhausting. And it takes away from time and effort spent with our core friends. However bartering or gifting with friends can be a very enjoyable social activity. What if you can easily and cheaply trade with players who are within your stable friend circle? A more free-form version of the same idea is to allow theft and betrayal so that trade tends to happen within high-trust circles.
Trade can be a specialized role that facilitates weak ties: Not everyone needs to trade with everyone to create large-scale, yet socially viable networks. Creating specialists trading roles that serve 150 to 500 people generates trade hubs that also serve as social hubs. These facilitate weak ties between denser, smaller friendship networks. Manage the density of trade hubs by culling those that dilute the 150 to 500 person sweet spot.
Pattern: Tying social metrics to business success
One of the great challenges of social design is that many business owners feel that it is an expensive extra. Should game designers play political games and show how social design drives business outcomes?
Find correlations with key business drivers
Split your population into higher trust and lower trust segments
Look for correlationships between trust and key business drivers like retention and monetization.
In general, you’ll likely find a very strong relationship. Intrinsic motivations, like social relatedness, are typically about 3X as strong a motivator than many of the extrinsic motivators found in more single player activities.
Use this correlations to justify additional investment in prosocial game design.
There are immense pitfalls that come from following this pattern. Profit motivated capitalism tends to be incredibly damaging to social systems design. See Dark Patterns below for examples.
Part 5: Dark patterns of economic design that sabotage prosocial play
Prosocial economics explicitly brings the tools of economics into social system design. And it promises to be an effective and scalable means of promoting societal values. This combination is a honeypot for bad actors. There is a future where the basic social technologies we’ve described in this paper will be used to create systems of immense evil, debasing the very aspects of friendship that we seek to elevate.
We’ve already seen some of these negative outcomes.
Facebook coopting social networks in order to sell unfiltered political advertising to the highest bidder.
China creating systems of social credit to survey and control those citizens who step out of a narrow range of acceptable behavior.
It is easy to imagine ideologically motivated governments, political parties and religious groups who co-opt the functionality of games to inject toxic tribal behaviors into the broader world.
Yet treating social systems design as a trade secret is also problematic. Again, the “alternative to good design is bad design.” To do good design, we need to grow a broad population of educated practitioners who are informed about both the craft and its negative outcomes. So that when things start going off the rails, we can identity and censure those who engage in dark social design patterns.
It is in the light of describing and enforcing ethical standards that we cover some of the darker patterns of prosocial economics.
Dark Pattern: Optimizing the system to improve proxy metrics instead of overall prosocial values
When a complex social phenomena (such as trust) is measured with proxy metrics, it obfuscates much of its expensive-to-measure nuance. This is exacerbated by the tendency to select proxy metrics because they are easy to measure, not because they are high quality proxies.
Subsequently, it is common for optimizers and balancers to start to mistake the proxy metric for the original phenomena. And as they make the proxy go up, they end up inadvertently damaging the hidden nuance of the original phenomena. Sadly, that nuance often turns out to be the real value we were trying to preserve and grow.
There are many examples of this:
GDP: In the real-world, you see governments optimize top-line GDP (gross domestic product). Over time, this proxy for economic growth and citizen well-being has become less correlated with these larger, more complex values. We see lower income segments suffering as small wealthy minorities accumulate the majority of newly generated wealth.
Togetherness in Steambirds: In our earlier example involving Steambirds Alliance, the togetherness variable was an easy-to-implement proxy for trust. As implemented, it depends on us checking if players are ‘near’ to another player when an enemy is killed. If we wanted to boost our togetherness values, we could simply increase the radius we check for ‘near’ players. With a large enough radius, the togetherness metric would hit 100%. However, despite the number going up, we’ve lost all insight into player ‘trust’.
Viral installs on Facebook: During the era of social network games, analytics teams optimized for increasing the virality of their games. ‘Virality’ was really a proxy for the complex social phenomena where a friend tells a friend about something they like and this trusted recommendation results in a highly engaged new player. In order to improve ‘virality’, social networks games began spamming friend lists with automated invites to games, often with minimal permission from the original player. These invites failed to trigger almost any of the important trust and reciprocation loops in the original phenomena. Instead, the spam damaged existing relationships and brought in low retention, unengaged new players. Pretty much all the games that seeded their audience with this distinctly low trust technique eventually failed.
Cross functional teams: Bringing multiple perspectives into the decision making process ensure that a single perspective does not dominate.
Holistic metrics reviews: Always return to the original prosocial pillars of the project and ask if your microdecisions and optimizations are still in-line with the holistic goals. What was the original intent of the proxies as they pertain to your prosocial pillars? Are you still measuring what you think you are measuring? It can be worth setting up a regular official review, but equally valuable is training the people making decisions in the field so they catch mistakes as they happen.
Player interviews: One method of capturing nuance is to talk to players directly. If you are only watching dashboards, you only witness what you are measuring. In depth qualitative interviews with key players uncover new trends and behaviors. What do they care about? What motivates them? What new skills or organizational techniques are they now using? You can then follow up with quantitative metrics gathering to understand the scope and impact of those behaviors.
Dark Pattern: Over reliance on extrinsic motivators
Motivational crowding is when a task that someone is intrinsically motivated to perform is instead encouraged with an extrinsic reward. As soon as the extrinsic reward ceases to be given, the person no longer wishes to do the task. Even if they were excited to originally do it for no explicit reward. The intrinsic motivation is said to be ‘crowded out’ by the extrinsic motivator.
Extrinsic motivators are much easier to put into systems. The game can dole out standardized rewards of commodity goods or currency and they can be triggered in a rote fashion upon the mechanistic completion of a well-defined task. For example, if we want to tell a person that their comment on a social media site was viewed and appreciated, we could add a ‘Like’ button and then report the total number of likes accumulated. We’ve turned a complex relationship into a tidy number you can watch ticking upward. Ding!
Intrinsic motivators are generally complicated and tied to an individual’s internal needs. Though intrinsic motivators are more effective, longer lasting and result in higher overall happiness of the person doing the task, they are far more difficult to design, measure and systematize.
The result is that designers tend to rely quite heavily on extrinsic motivators. And in the process, inevitably damage our intrinsic motivations. This is highly problematic in social spaces, since social interactions tend to be intrinsically motivated and involve nuances unique to each individual relationship. Whoops.
In this era of modern computation, there is no reason why we can’t be far more targeted and contextual with our incentives. By tracking where each person is on their personal journey through their game progression, through their acquisition of friends, through their micro actions we can create personal models for what they might desire.
Stop designing for populations of average players and start designing for the intrinsic motivations of the individual player. Even small shifts in this direction, such as facilitating activities based off the state of a player’s direct friend network, can have large positive impacts on engagement.
Scope of metrics and their impact as extrinsic motivators
Social metrics such as a ‘Like count’ can quickly turn into extrinsic motivators if you aren’t careful. Carefully scope how your metrics are revealed to minimize negative impacts.
Internal: In general, most social metrics should be Internal, metrics that are only shared with the internal development team. This eliminates the dangerous feedback loop in which a person attempts to influence their own metrics.
Private: Slightly riskier is a Private metric where you share a person’s information with just that person. This creates a feedback loop but it is limited to only that individual and they have full control over what they do with the information. This is especially important for sensitive information around reputation or facts that could be damaging to share without a foundation of trust.
Group: One layer out from the individual are trusted friend circles or affiliation networks. We start to see social metrics generating politics, censure and other group dynamics. These are intense feedback systems that can result in unexpected results. This is known colloquially as drama. At this scope, we also start to see new intrinsic motivators based off status start to arise. Status-driven motivators can start to offset some of the motivational crowding. Of course this only works for high status individuals in the group. Low status people who don’t see their public metrics move respond as if they’ve been shunned. This negative outcome is exacerbated by social anxiety.
Public: The most dangerous metrics are public ones that are shared broadly. We get all the drama of group interactions, but we also get in-group and out-group competition. Various tribal groups use sensitive information to engage in hate mobs, griefing and other forms of abuse. Popular status seekers, especially those with narcissistic tendencies, thrive in these spaces. Use public social metrics with immense care.
Questions to keep in mind
There’s no clear fix for this issue. Instead, I try to keep myself honest by asking several questions periodically.
Are you applying an extrinsic motivator to something that players would do of their own volition? Many times we apply rewards to activities out of habit. Pause for a moment. Does this social interaction really need a flashing reward screen with a loot drop of crafting materials or currency?
Can you build the context within which intrinsic motivation can take over? Instead of mechanically telling players to do activity A for reward B, you can instead provide a space to do Activity A and highly visible affordances. Give players a small amount of room to stumble upon the activity. And pursue it if they want. Animal Crossing is a lovely example of this sort of constrained small space and intrinsically motivated activities.
Dark Pattern: Replacement of prosocial values with selfish values
The most likely source of corruption of a prosocial economic system is when it managed by an unreformed capitalist. An executive who believes in the selfish nature of humanity will tend to replace the core prosocial values with processes that are shortsighted and profit motivated.
Economists (and capitalists who love economics) tend toward evil
Those that practice economics — and to a degree modern American capitalism — are heavily invested in an implicit system of self-centered moral values. A well-documented phenomena is that economists behave more selfishly than other professions. They are less fair, less loyal, less cooperative, more prone to deception, and give less to charity. This appears to also impact executives who use economic framing of problems.
In part, this seems to be due to the field of economics attracting selfishly motivated people. But it also appears to be the result of indoctrination. The repetitive doctrine that humans are best modeled as selfish rational optimizers creates a set of selfish social norms that practitioners consciously or subconsciously follow. The act of studying economics makes you a morally worse human-being (by most definitions of morality.)
There is another possible cause for this selfish behavior, which is economists’ high exposure to commercial systems. The presence of currency itself, and the tracking of it and focusing upon it, seems to lead to rationalizations that justify selfish behavior. We see this in particular in games as a dimension of the above dark pattern, reliance on extrinsic motivators. Pure exposure to extrinsic motivation systems, of which accumulation of currency is one, seems to bend human behavior toward norms that justify maximization of that accumulation. It is possible that the persistent high exposure to game currency — and as we have stated, almost all games have currencies and tangible economies of some kind — has the same effect that exposure to economics has on economists.
Values as identity
These values are embedded at the level of personal and tribal identity, and so in groups they become naturally amplified. When challenged, the result is a blunt dismissal of any information that disagrees with the existing world view and a re-entrenchment in existing beliefs. One merely needs to read the responses to some of the studies on selfishness in economics to realize this is not an open-minded, self-reflective group. (My favorite is that claim that economics is perfect, it is merely all other fields of study that mistakenly train up altruistic, prosocial citizens)
A clash of values
When worldviews clash, those with the more power wins. A powerful executive, indoctrinated in the ways of selfish capitalism, is very likely to dismiss the prosocial value at the heart of social system design. A very difficult argument to win. Prosocial design presupposes an altruistic model of human behavior that has long been scrubbed from the selfish predator’s worldview.
We’ve seen this first hand with companies like Zynga, where capitalist managers methodically and deliberately optimized delightful games about creativity and sharing (Farmville) into viral advertising engines that actively degraded relationships. Even in the face of their market crashing, at no point did they stop and question their selfish worldview. Instead they doubled down on burning out more players to maximize revenue extraction.
Be explicit about key prosocial values. State prosocial design goals as key product pillars that are inviolable. Have someone who understands prosocial systems own and enforce these across the entire company.
Tie economic value to the maintenance of prosocial value. Get the profit-minded forces at the studio on board with a win-win partnership. Use facts like intrinsic motivations being 3X as powerful as extrinsic motivations. Or the crowd-out effect of extrinsic motivations damages long term LTV. These can turn the selfish desires of executives into support for prosocial design. Align prosocial design with smart business design.
Don’t put business in control of prosocial systems: Create an organizational firewall between those handling the prosocial systems and those directly driving profits. Acknowledge that your short term focused business teams may not have the values, goals and mindset to properly grow and manage business critical social systems.
Hire prosocial executives that know the value of these systems: Instead of fighting a losing battle with entrenched executives who have a long history of fetisizing selfish economic behavior, seed new teams with strong leadership who already buy into the mission of building prosocial games.
Ethics standards for social systems designers: A longer term dream is to create ethical standards for social systems designers. Perhaps in the future we could build training programs for this deep skill set. And bind trained students to a set of ethical standards. There would need to be some form of censure as well if lines were crossed.
In conclusion, we have described:
All games with resources have economies.
Economies that do not consider their end aesthetic outcome devolve into antisocial patterns. Specifically toxicity and loneliness.
Real world economics often undervalues or ignores prosocial behavior. It is challenging to apply directly to games.
There are a handful of prosocial economic patterns we can use as designers.
Economies, being motivational systems, are inherently subject to exploitation and dark patterns.
Summary of Patterns
Prosocial mechanical and economic patterns identified in this paper include:
Measuring the Unmeasured
Measuring trust (quantifying social capital)
Positive sum resources
Integrating social metrics with business success
This paper is intended as an initial exploration in the domain of prosocial economic game system design. Much further work is needed to explore, codify, and test these patterns and ones that may be discovered after.
Patterns that we identified but have not built out in this paper include: 1) group leveling, 2) friendship resource (differentiated resources), 3) incentivizing generosity, 4) nurture play, and 5) expressive orthogonality through fashion.
Further areas of interest uncovered by our preliminary exploration include:
Prosocial economic patterns: Further expanding within the macro-patterns of prosocial design patterns.
Positive sum design: Expanding the design patterns for non-scarce (positive sum) resource mechanics (game design without scarcity).
Public goods design: Expanding design patterns around managing public goods, especially via decision mechanisms
Therapy: Leveraging disordered personality behavioral archetypes and corresponding treatments for CBT-based (as one example protocol) game progression systems.
Managing social progression systems: Scaffolding social skill development, and differentiating social skill development from meaningful relationship cultivation.
Transfer to non-game environments: Transferring skills and social capital experienced in game environments outside the game environment (crossing the membrane).
Dark patterns: Further exploring, and codifying, dark design patterns in extrinsic motivational systems, and their consequences;
Education: An open protocol of transparency and education regarding game-based motivational systems and economy design.
Ethics: Ethical rules for social systems designers as well as institutions who help promote those rules.
Appendix I: Towards an action-based framework for mitigating loneliness
A widely used instrument for detecting loneliness is the Roberts UCLA Loneliness Scale. First developed in 1978, it is estimated to have been used in 80% of scientific research studying loneliness, and has been found valid by multiple meta-analyses — so it makes a good starting point.
The original 20-factor Loneliness Scale has been condensed into smaller sets including the RULS-8, RULS-6, and RULS-3. We are primarily referencing the 1996 20 point scale, and distill from that scale some concepts sometimes referred to as “dimensions” of loneliness. These dimensions have been studied in medical research, but for our purposes we are proposing a conceptual framework of loneliness dimensions most relevant to game behavior:
Exposure: Relating to social vulnerability, exposure is feeling unsafe because one is alone. This often relates to the involuntary nature of the exposure, as contrasted with, for instance, solitude, which is a positive feeling of isolation rooted in its deliberate and voluntary nature. Exposure is a broad category of loneliness relating to index measures “I lack companionship”, “I do not feel part of a group of friends”.
Ostracization: An experience of social punishment, ostracization is the feeling of being “left out”, and especially being “shut out”, of one’s valued social group.
Shyness: A withdrawn state resulting from feelings of social isolation. The feeling that it is very risky to disclose oneself to others; a feeling that others will not understand you. Relates to “I am unhappy being so withdrawn”.
Unfit/Outsiderness: A feeling that one is around others but does not belong there. Belonging is a broad concept (discussed below), but “unfitness” relates to being in the presence of others with whom one does not feel connected. Relates to measures “People are around me but not with me”. Importantly emphasizes the presence of others who are not of one’s belonging group, with the absence of one’s belonging group.
Loneliness causes stress in humans broadly, relating to feelings of vulnerability, and can also provoke scarcity mindset, in which a host of negative outcomes occur. Scarcity mindset is a stress-induced “tunnel visioned” state that causes short term thinking associated with long term net negative outcomes.
Kinds of loneliness
As a creative empathy exercise, it can be helpful to identify distinctive, separate feelings of loneliness for which there aren’t English words:
I have friends but I can’t count on any of them
I have close family but they don’t understand me
There are things I can’t share with my friends
There are things I can’t share with my family
There are things I can’t share with my spouse
I have good friends but our group doesn’t have purpose
I feel lonely and socially exhausted around my friends
These limited examples illustrate some of the complexities of loneliness, which represents a rich artistic space rife with subtlety and inner conflict.
Structurally, it can be helpful to think of two large categories of loneliness:
Emotional loneliness: Lack of an attachment figure
Social loneliness: Lack of social connection; social vulnerability; social isolation
It is important to note that not all loneliness is purely social or purely emotional; these are two separate dynamics that combine to produce the emergent sensation of loneliness. Emotional loneliness in particular is especially tractable in digital/fictional experiences; reading a book or taking care of a fictional animal can assuage emotional loneliness, even though these are solitary activities.
Amongst the more complex category of social loneliness, we can identify sub-categories as well:
Affinity (sharing interests)
Recognition (feeling known deeply by others)
Belonging (feeling accepted): These can be split into Acceptance (feeling wanted and not judged) and Usefulness (filling a distinct need/role in one’s group)
Companionship (feeling the presence of other social creatures)
When we are talking about prosocial game design, we are, in part, talking about game design systems that address the social pain of loneliness. By dividing loneliness up into its distinct constituent categories, we can more accurately aim experiences at assuaging specific target areas.
What game designers need to know about loneliness
From a game design standpoint, there are some important high level takeaways:
There are multiple types of loneliness; it is not a single phenomenon;
Loneliness and isolation are distinct psychological problems;
Loneliness can be clustered within orthogonal sub-categories that must be independently addressed;
Loneliness is best thought of as a specific kind of social pain.
Prosocial design has a lot of interesting tools for tackling social loneliness. We have fewer tools for tackling emotional loneliness though this is a fascinating area of further investigation.
Appendix II: The economics design lens
Economics is one of many potential lenses, or perspectives for understanding a game systems. As a designer, it is critical you can swap out lenses for examining a problem as needed.
For example, you can take a system like player chat and look at it via different lenses and learn something new from each.
Seen through an economics lens: Use the tools of economics to assign values to relationships and track the time payments back and forth between chat agents.
Seen through a psychological lens: However, we could just as easily look at chat from a psychological perspective and gain a set of insights that are impossible to capture with a purely economics lens. The emotional tone of a snarky response for example can be challenging to model with our simplistic set of tokens and transforms.
So what is the economics lens good for? It helps to think of the lens of economics in game design as having a couple basic superpowers. These end up also being its core weaknesses.
Economic super power: Analysis and balancing
Almost any game with a heavy systems focus benefits from using economics to balance or tune the systems to achieve a specific aesthetic outcome. There are several key steps in this process that build upon one another.
Definition: The economics of a system become visible the moment you start defining the exact tokens, source, sinks, etc. What you find out depends entirely on the quality of your definitions. And poor definitions result in weak insight.
Economic analysis: Once you’ve definite the components of the economic system, we can start interrogating why something is happening. The type of analysis you can do is limited to answering questions about resource flows and transformations. There are huge swaths of the gameplay experience that are hidden or only observable through secondary effects. For example, economic analysis can say little about a beautiful experience, but it can track the price and availability of that experience.
Balancing: Finally, an economic lens allows us to ask what-if scenarios, adjust our various defined economic components and then analyze and observe the results. You are always balancing in order to some overall aesthetic goal (In the Mechanics Dynamics Aesthetics sense of the term).
Small errors accumulate
Now, the clever reader will notice that the balancing step is built upon an unreliable stack. If your definitions are incomplete, your analysis will be flawed. If your analysis is flawed, the initial balance ideas will be impossible to verify. This is particularly challenging when your changes alter the very nature of the virtual world you a measuring. There less in common here with natural science than might be hoped. The iterative act of balancing an economic system in a virtual space can quickly turns into feedback loops where small errors accumulate.
Add in poorly modeled humans as key decision drivers and you can very easily design something that is a bit of a mess. Economies in games are often prone to inexplicable and unexpected exponential failures. We call these disasters by different names (ex: Mudflation, grindy, OP) but they are often failures of economic balancing.
The predominance of toy-like economies
So we punt and build toy-like economies that are trivially understandable (as is the case with most single player games). Or we build systems that are stable short term and a spiralling disaster only if left unmanaged (most multiplayer games). For multiplayer systems, we continuously micromanage them into some rough stability using god-like powers to shift the virtual world’s physics if things get too far off.
The bigger lesson here is that in practice, economic tools are essential yet unreliable design tools. Especially at scale. So we build systems that compensate and can be balanced despite the flaws in our tools.
Economic super power: Efficiently generating value through trade
Perhaps the single most meaningful insight that economics has added to the world is that trade generates material value for society at scale. The orthodoxy of economics may have poisoned a richer discussion of the topic, but kudos for the acknowledgement the historical practice and clarifying why trade is important.
Trade in games is mostly studied in the context of multiplayer games with player-to-player exchange of virtual items. For a primer, you should read Virtual Economies: Analysis and Design by Lehdonvirta and Castronova. Though designers have learned many lessons over hundreds of MMOs, it still remains a niche field of practice. In this modern era, many hyper focused, metrics-driven teams try to stamp out trade entirely due to the unmanageable chaos it creates. Once you introduce capitalism into your toy economy you’ve opened Pandora’s box of design challenges, both economic and culture.
The basics of trade
The basics go back to Adam Smith.
Person 1 has access to resource A. But they know they really need resource B.
Person 2 has access to resource B. But they know they really need resource A.
So they trade with one another! Now they both have what they actually want.
The magical bit: This was a positive sum exchange. Player 1 is happier and so is Player 2. Value has been generated from trade.
A drive for more efficient trade
This sort of basic barter certainly works, but the logistics are complicated to arrange. So we introduce an intermediate currency and use that to value both resource A and B. Now each person can just set a price for goods they are willing to buy or sell. As long as there’s a cheap way of sharing prices, any person can sell their low value goods to someone who values them more. And then take that excess money to buy the stuff they really want.
So why is this so interesting?
Prices are set locally. The local agents determine what they value, so in theory you can just put a bunch of independent agents that know their own needs together in a common trade area and you’ll start to see market dynamics. The setting and sharing of an agent’s prices are a decoupling mechanism that allows the system to create somewhat scale free networks.
With relatively little managerial oversight, a vast number of people can trade with one another efficiently. You do need some institutional protection or bad actors can start to sap profits through theft and extortion.
Each iterative trade in turn generates enormous value across many of those participating in the market.
That excess value is now possible to redirect into things like culture, leisure, research and moving beyond mere survival.
This process, when the right conditions exist, can be explosive. Huge amounts of material and human resources gain explicit value and are efficiently send zipping around in complex, somewhat self-organizing system that radically transforms everyone and everything involved. Capitalism, writ large, has according to some metrics, resulted in some of the greatest increases in human health and stability history has ever known.
The inevitability of trade
Trade has a degree of inexorable social physics to its emergence. Most large scale societies develop it in one form or another; though rarely to the degree of modern capitalism. We see this capitalist explosion in multiplayer games all the time. The basic requirements for barter seem to be:
Players can exchange goods,
Those goods are differentiated and randomly distributed
Players can negotiate relative value with one another.
Once barter is in place and the society is stable enough to create community standards, players develop an emergent currency (usually some easily tradable item with a stable supply) This is then used to facilitate efficient trade networks. In mere weeks or months there are merchant classes, black markets, trade, commodities, trust networks and more.
Another perspective (a lens!) on economics is that it is a memetic virus that transforms a society and distorts it to fit the functional needs of the virus as well as fostering the cultural values that help the virus spread and thrive.
Game developer superpower: Economic design tools available to game developers
Many of the practical issues that weigh upon real-world economics impact game developers less. Game developers benefit from the following factors:
Large human populations: A successful online game has many thousands of players. We can run experiments with real people without as much reliance on bad models or small samples.
Sources of new players: We have access to new players who are more of a clean slate for testing out new ideas.
Rich data collection: Our analytics can capture any economic or social interaction inside the game. Including rich historical streams of individuals or populations.
Control over most laws of nature: We have full control over sources, sinks, transforms and any associated incentives. We can change the world to fit the model almost as easily as changing the model to fit the world. Players, math and time are still out of our immediate control.
Less politics: Online games are benign dictatorships with voluntary membership. Though there are some checks and balances on performing radical economic experiments, there immense leeway to make changes.
Appendix III: What does economics say about altruism?
Most economics theory is based off the idea of a rational, self-serving actor. Economic is not wholly ignorant of altruism. It merely is treated as a series of side theories that are not broadly integrated into mainstream economic models or policies. It is worth mining these theories to see if any of them are applicable to the design of prosocial economies.
What is altruism?
In economics, altruism can be defined as investment in public goods. These are shared resources or investments that benefit multiple people, not an individual owner.
Note that altruism and prosocial behavior in trusted relationships are not exactly the same thing. Altruism does not require trust, merely a shared public good. Though shared relationships at the heart of prosocial systems are almost always a public good within the local context of the relationship.
Onto the theories. We’ll start out with the earliest and most wrong theories and then progress to ones that slowly incorporate more experimental support.
If people are rational actors, when it comes to public goods, selfish people should act as free riders. Assuming most people are selfish, this would result in public goods being under provided for because most people free ride on the irrational contributions of a few. Examples of this include
However, people free ride less than expected. They are not purely homo economicus, the selfish man. Cases where they over invest according to self-interest theories include
Contributing to open source software
Theory: Incentivized prosocial behavior
Not willing to let go of the belief that people are inherently selfish, a variation of the self-interest theory is that people contribute to a public good are in fact getting paid. It is just in the form of non-obvious currency such as prestige. In practice, this doesn’t hold up since people donate charities anonymously.
Theory: Pure and impure altruism
We now get into outcome-based prosocial preferences. What if people inherently enjoy seeing the well being of others, so they contribute to public goods? Imagine we gain internal utility (a ‘warm glow’) by helping others, so helping is intrinsically rewarding.
This also doesn’t match observed results. First, even when others are doing well and don’t benefit, people still donate. Second, such an intrinsic motivator would be a stable source of motivation. No matter what we should keep donating if there is continued need. But prosocial behavior decays with repetition. And people have this distinct tendency to punish the behavior of others. Which is a bit inconsistent with a purely altruistic motivation.
Theory: Inequality Aversion
What if we just hate inequality? Imagine that one’s relative standing in the leaderboard of income distributions drives people to reward those less well off and punish those more well off.
This one doesn’t explain a lot of nuances about when and how people punish and reward others. Especially across different cultural contexts.
Theory: Reciprocity and Conditional Cooperation
Okay, what if who we are interacting with matters? Now the theories start to include some basic social psychology like reciprocity in their human models. And some interesting findings start popping up.
Norm enforcement is intrinsically motivated: Expensive punishment of free riders is behavior that people perform even in the face of repetition. Almost all extrinsically motivated behavior drop off with repetition. This is a key finding since it suggests that enforcement of social norms is an intrinsically motivated behavior.
Altruism depends on perceived laziness: If someone sees a recipient as lazy, they tend to reduce donations to them.
Reciprocity drives behavior: If you give a gift, the other party will often give one back. However, intentions also matter. Why someone does something impacts how the other party reciprocates. This is a big effect that also continues with repetition.
These observations also lead to the prediction that if more people act prosocially, an individual will be more likely to act prosocially. For example, one’s donation depends on the donation of one’s reference group. A 10% increase in donations by the reference group results in a 2-3% increase by the individual. So people are conditionally altruistic based off the social norms of the group.
Theory: Self-identity theories
A person ends up identifying with a reference group. And they’ll be more prosocial if two factors are true
The reference group thinks the action is good
The action is a valuable signal of the person’s good traits as determined by the social norms of the group.
Theory: Frame effects matters
Now we start moving away from universal models of human behavior and begin to dig into the question of how context (aka the institutional environment) impact what someone decides to do. I think of this as economists discovering the importance of level design. There are a large number of studies on ‘frame effects’.
Do you have knowledge of free riding by others
Do free rider know they are being observed and by whom
Can you punish free riders
How strongly can you punish free riders?
How were resources earned?
Is the recipient a charity?
Is the recipient a close friend or relative? A general group or a specific person?
Is this your ingroup or an outgroup?
Are the bad circumstances you are alleviating due to bad luck or poor choices?
Did you form an agreement with the other party? Even if the agreement has no binding value, people rarely break them.
Additionally, the type of communication you have with the other benefactors of the public good matter. At this point we are starting to get really close to friendship formation and intensity as an accelerant for trust accumulation.
Do you get to talk to the other person?
Was it face-to-face? If so that results in a strong impact on altruistic behavior. .
Was it via a computer? If so there’s a much weaker impact on altruistic behavior.
Framing is another name for much of what we do as game designers as we set up contexts for players activities. There’s a wonderful exploration of reframing economic activity using game worlds in the book Stealing Worlds by Karl Schroeder.
Theory: Monetary incentives in the world of frame effects
Finally we roll all the way back around to extrinsic motivators. But this time we are looking at ways that the system designer can create frame effects that alter an individual’s behavior.
If the systems makes an action cheaper or easier or slightly incentivized, the intervention can increase prosocial behavior.
However if you increase benefit too much extrinsic motivation ‘crowds out’ intrinsic and behavior drops.
If monetary giving goes up, so does giving of time (complementary goods)
Reliance on extrinsic motivators selects for selfish people.
Theory: Heterogeneous populations
At some point in all of this, someone raised their hand and says, “But what if different people engage in different strategies?” Individuals are heterogeneous; some tend to use one pattern of behavior while others use other patterns. A community is an ecosystem of agents, who depending on local conditions, take on different social roles.
In some tests: 23-30% of the population is always acts selfishly. No matter what. But 50% operate conditionally and are likely to behave altruistically if the right conditions exist.
Presence of a reciprocal person causes other conditional people to reciprocate. Thus shifting the whole dynamic.
Maybe teaching ethics helps create less selfish individuals. Economics hasn’t studied this yet. (Students going into economics are more likely to be egotists already)
Amy Jo Kim, Chief Executive Officer at Shufflebrain
Crystin Cox, Principal Program Manager at Microsoft
Daniel Cook, Chief Creative Officer at Spry Fox
Erin Hoffman-John, Lead Prototyper at Google
Isaiah Cartwright, Game Director at ArenaNet
Kyle Brink, Director of Production at ArenaNet
Link Hughes, Game Designer at ArenaNet
Many of the problems associated with making an MMO, a Massively Multiplayer Online game, come in large part from the very first term: “Massively”. An MMO is notably tricky to build due to technical issues involving server scaling, as well as design issues involving scaling economics, politics, level design, pacing, persistence, and progression. A rule of thumb is that development costs grow exponentially as the number of players increases, but for many years, there’s been an unquestioned assumption that bigger player numbers are inherently better and therefore worth pursuing.
Yet we see clear counterexamples. Many early MUDs (Multi-User Dungeons) involved populations of dozens-to-thousands of people and still have vibrant communities to this day . Multiplayer Minecraft is wildly successful, despite its reliance on relatively small, instanced servers. And many modern hit games, like Fortnite, are online games that successfully limit their focus to matches of 100 or less.
What are the critical design lessons from these smaller online games—and how can current research and understanding of social psychology help make sense of those lessons? We combined our decades of experience designing social systems for online games and a deep dive into current academic research to arrive at a set of best practices and common pitfalls.
What we’ll cover in this paper:
What we can borrow from social psychology
An overview of friendship formation
Dunbar’s Layers and the constraints they place on social systems design
Social group and the constraints they also introduce
Big design insights
Opportunities for fulfilling the social motivations of players
Borrowing from social psychology
When researching what it meant to make human-scale systems, we found several key concepts from social psychology. Each provides a set of constraints for social design. Social game design operates within the physical and mental constraints of the human animal, so it pays to understand these constraints and build them into our designs.
A friendship is a single social bond between two people. Friendship formation is a distinct process involving proximity, similarity, reciprocity, and disclosure.
An individual has a highly structured distribution of relationship bonds. People tend to have a maximum of 150 total friendships , including 50 good friendships, which include 15 best friendships, which, in turn, include 5 intimate friendships. This web of relationships can be modeled as an egocentric network with the individual at the center. This paper focuses primarily on the implications of Dunbar’s Layers for human-scale social design in online games.
A social group of is a collection of people brought together for a shared task or interest. Groups contain multiple overlapping individual networks. The performance of the group, as a whole, is dependent on how the friendship bonds across the entire group are leveraged to accomplish the shared activity.
At the most basic level, human-scale game design is about creating strong relationship bonds between individuals. Most game populations will start out with weakly-bonded individuals. You’ll need to create activities, incentives, spaces, and social structures that actively build friendship in order to enable even the most basic of trust-based activities.
This section is a brief overview. For more detailed discussion on this topic see the 2016 Project Horseshoe paper on game design for building friendships.
The basics of growing friendships
Friendship formation requires 4 key ingredients:
Proximity. Being close together to one another encourages frequent serendipitous interactions.
Similarity. Players will generally be more likely to become friends if they perceive one another to be similar.
Reciprocity. Players must engage in escalating back-and-forth interactions in order to negotiate shared social norms.
Disclosure. At higher levels of friendship, there needs to be an opportunity for safe, consensual, intimate sharing of weaknesses.
You can take any two players, put them together in matches for hundreds of hours, and if the above criteria are not met, they are unlikely to become friends. Naively tossing bodies at one another is not efficient social design.
The micro-design of social systems is all about reciprocation loops
As a designer, you specifically have to build opportunities for consensual reciprocity into your game loops. These look like the following:
Opening. A person performs an opening action that a second person observes. This action has a cost in terms of time, investment, skill, and other resources. For example, Player 1 asks a question in open chat, which costs time and social capital.
Opportunity. An opportunity is created for the second person to respond. For example, Player 2 sees the question and can answer in the same chat.
Response. The second person performs a responding action that acknowledges the first person. This also has a cost. For example, Player 2 offers an answer to the question in chat, which also costs time and social capital.
Acknowledgement. The first person acknowledges the second’s response and the loop is now complete. For example, Player 1 thanks Player 2 for answering their question in open chat.
Escalation. Either during the Response or the Acknowledgement stages of the loop, a person can escalate by opening up a new loop or prompting additional response. This is an opt-in act. For example, Player 1 asks for additional details.
Rejection. Either during the Response or the Acknowledgement stage, a person can either not respond or respond inappropriately, which also collapses the loop. For example, Player 2 mocks Player 1 instead of answering their questions.
Link loops together in an escalating structure
Friendship is a long-term process. Each reciprocation loop may take seconds initially, but you need thousands of linked loops to build a robust friendship.
Create low-cost loops with low rejection costs for early relations.
Create higher-cost loops for later term relationships.
Build space inside the later loops for expression and definition of the personal relationship between two players.
For example, friendships in an MMO tend to start out with parallel play, where two people simply see one another’s name while fighting monsters in the same area. This then escalates to helping one another; a heal spell, an emote of celebration, a dropped item. The two players may start chatting in order to take down harder monsters; they may also friend one another and start talking more about who they are and what they are interested in.
At each stage, interactions take increasing time and effort. And involve richer communication. Each micro-loop is not very expensive, but over long-term repetition of many such loops, the relationship accumulates meaningful amounts of trust.
Design these systems with the same rigor, care, and eye for economic balance that you’d put towards a combat or progression system.
Design for consent
Almost every stage of these reciprocation loops involves consent. Each party must consent to both starting, continuing, and escalating the relationship. At any point, it is totally fine for one or both parties to pull away, either to slow down or move onto some other relationship opportunity.
In the context of Dunbar’s Layers, there’s a limit on the number of people an individual can have in their lives. The process of building friendship is also the active process of curating relationships that are healthy and mutually satisfying. When players actively and enthusiastically consent to engage in your reciprocation loops, you’ll find that the relationships you build in your game are more authentic, last longer, and ultimately provide more value to your players.
An individual organizes their friendships by strength of their one-to-one bonds. They have close friends they turn to in times of crisis and more casual friends, with whom they interact with less frequently. Social psychology has been studying these friendship networks for decades. One of the more reproducible findings is the existence of strong limits on the number and strength of bonds an individual can have with other humans.
Robin Dunbar is an anthropologist who, in the 1980s, posited that a human can have up to 150 meaningful relationships, based off his investigations into primate social brain structures . When others attempted to verify this prediction, they found that “Dunbar’s Number” kept coming up in long-lasting groups in the real world. It’s been replicated across a huge number of domains including businesses, religious organizations, military groups, and, of course, MMO guilds.
Multiple layers, not a single number
However, as researchers dug further into the data, they noticed additional stable clustering at lower numbers of connections. These smaller clusters were part of a person’s total of 150 relationships, but involved much stronger bonds.
Visualization of Dunbar’s Layers. Each block represents time to build one relationship in that layer.
Dunbar’s Layers, as these smaller clusters are known, are generally organized as follows:
1.5 people: The intimate couple or the individual.
5 people: Intimate friends or family. People you can call in a crisis.
15 people: Best friends. People who you can ask for sympathy.
50 people: Good friends. The majority of regular social contacts and, by extension, all of one’s emotional and economic support .
150 people: Casual friends or acquaintances.
Note that each layer is cumulative and contains the previous layers, so your best friend layer contains your intimate friends layer as well. A common confusion is to think you have 5 intimate friends AND an additional 15 best friends, etc., but those 5 intimate friends are part of your 15 best friends budget.
These numbers are averages and, in reality, describe tight ranges. In practice, different people have different degrees of social needs and relationship-building capacity. For example, many men average 3-4 relationships that they would consider intimate friends or family, while many women average 7-9 such relationships. Some people, known as “super-connectors,” have upward of 200–250 meaningful friendships.
With larger data sets, we’ve discovered these relationships layers also extend past actual friends.
500 people: Nodding acquaintances.
1500 people: You recognize their face, but that’s it. 2000 faces seems to be the absolute maximum that a human can recognize and when you learn a new face, you drop one of the other faces you’ve memorized.
Implications of Dunbar’s Layers
On first glance, Dunbar’s Layers are a mere curiosity. However, they fundamentally shape how people socialize. The following are aspects of Dunbar’s Layers worth knowing about before you attempt to use them in a design.
Dunbar’s Layers are egocentric networks
Visualizing the innermost Dunbar’s Layers as an egocentric network. Note all connections are from the perspective of a single individual.
An ideal way to visualize Dunbar’s Layers is as a network of connections, not as separate layers, per se. In research, this is known as an “egocentric network.”
Put a given individual at the center of a nodal network.
Then map out bonds going directly to that individual. You’ll end up with an average of 150 meaningful relationships connected to the individual.
Some bonds between the individual and their friends will be stronger than others. These bonds map onto Dunbar’s Layers. For example, a person will have an average of five strong bonds.
There are several different ways egocentric networks can be used in analysis of individual relationships:
Dyads: The relationship between any two individuals involves two connections, not one. Each person has their own perception of the connection’s strength. It is possible—and, in fact, common —for these perceptions to be unequal.
Triads: Friendship networks can be analyzed by looking for triads—groupings of three people with at least two relationships between them. The strongest network structure is a “Triadic Closure,” wherein all three three individuals share mutual friendship bonds of equal strength.
Strong Ties: When a person has a direct relationship with another person, it is known as a strong bond. Strong bonds are key to meaningful relationships, support networks and overall happiness.
Weak Ties: When a person’s friend has a relationship with another person, but the original person does not, it is known as a weak bond. Weak ties are critical to connecting independent social groups and are particularly important for the functioning of large scale economic and informational systems. Weak ties also populate the 500 and 1500 person layers. When we start discussing social groups, weak ties become a very important concept.
Super-connectors: Some individuals have substantially more than the the typical number of connections. Known as “super-connectors,” they end up acting as hubs that connect disparate friend networks together.
Close friendships have a strong influence on quality of life
Overall, having a deep friend network has an immensely positive impact on your health and happiness.
High quality, high intensity relationships are positively correlated with increased life satisfaction .
Depression is lower overall in individuals with rich friend networks .
On the flip side, toxic relationships have an outsized negative impact on mental and physical health. Something to think about when we deal with trolls in our games .
High trust relationships take time and the right context
Building friendships takes many hours of interaction.
The time required to build a single friendship bond :
Casual Friend: 40-60 hours
Good Friend : 90-110 hours
Best Friend: >200 hours
If you meet with someone for 1 hour each week, it will take roughly a year before you consider one another even casual friends. Friendship formation is not a cheap activity.
Maintaining relationships takes less effort. Three key variables here are kinship, gender and frequency of interaction. Kin bonds (bonds with family members) require less maintenance than non-family friendship bonds and do not seem affected by distance. Men tend to affirm bonds by participating in activities together, while women tend to talk with another. Higher strength bonds needs more frequent renewal than lower strength bonds.
Casual friends meet up at least once a year.
Good friends meetup up once every 6 months.
Best friends meet up once a month.
Intimate friends meet up at least once a week.
You can’t beat the system
One way of thinking about the constraints suggested by Dunbar’s Layers is to imagine you have a budget of cognitive resources that can be spent on relationships. The physical limits of your human brain mean that you only have enough mental budget for a total of roughly 150 relationships.
Humans have developed a few tools that have expanded our ability to organize into groups well past our primate cousins—most notably language—but also large-scale systems of government and economics. In the early 2000s, people assumed that new technologies like online social networks could help break past Dunbar’s Number; by offloading the cost of remembering our friendships to a computer, we could live richer, more social lives, with strong relationships to even more people.
We now have copious data that this is not the case. Studies suggest that there’s still a limited budget of cognitive resources at play and even in online platforms we see the exact same distribution of relationships .
If anything, social networks damage our relationships. By making it possible for us to cheaply form superficial relationships (and invest our limited energy in maintaining them), such systems divert cognitive resources from smaller, intimate groups out towards larger, less-intimate groups. The result is that key relationships with best friends and loved ones suffer. And, unfortunately, it is the strength of these high-trust relationships that are most predictive of mental health and overall happiness .
What is a social group?
A social group is a set of individuals labeled as being in a group. This is inherently a fuzzy concept, since the true structure thereof is an overlapping network of egocentric networks, partially-negotiated social norms, and ever-shifting relationship bonds.
There are three dominant perspectives on what makes a group.
Social Identity perspective: “I feel like I’m part of a group.” An individual can self-identify if they are part of a group. By doing so, they start practicing the social norms of the identified group. This is the perspective that gives birth to either imposter syndrome or a feeling of belonging.
Self-categorization perspective: “I feel like you are part of a group.” Someone looking at the behavior of other people can identify if others are behaving as part of a group. By doing so, they treat those people as if they operate using shared social norms. This is the perspective that gives birth to stereotypes.
Social cohesion perspective: “We act according to shared social norms.” A set of people that act in similar manner across a variety of social variables is a group. Those variables include:
Shared goals. Are we working towards the same purpose?
Roles. Who does what?
Status relationships. Who has power?
Norms. How do we work together?
Sanctions. What happens when norms are violated?
Additional factors that can be used to determine group cohesion include:
Group size. How many people are in the group?
Group trust. How strong are the bonds between individuals in the group?
Group stability. Does the group come together for a short period of time or is it a stable, persistent entity?
The social cohesion perspective proves the most design insight, so we’ll be referencing it for the rest of this discussion.
Common groups sizes roughly align with Dunbar’s Layers. However, these are not identical concepts. Social groups can contain friends of varying trust levels. You could have a small group composed entirely of strangers. whereas a 5-person intimate friends layer is, by definition, an individual’s closest set of friends.
Small friend groups
These are some of the most common task-oriented groups to form. Non-kin, task-focused groups of these sizes often dissipate when the task is complete. Small groups are, however, able to attain the highest strength of social bonds, usually focused on key family relationships.
Pair. 2 people
Small Group. 5 ± 2 people
Medium group. 15 ± 6 people
Large social groups
These are the largest-possible friend groups. Example groups at these sizes include a guild, shard, or map in an MMO, a mid-sized company, or a social organization in a university.
Band. 50 ± 18 people
Clans. 150 ± 50 people
Huge impersonal groups
These larger groups are composed of smaller friend-based sub-groups. However, due to their size being larger than Dunbar’s Number, it is impossible for them to engage in very high-trust activities without additional systems like hierarchy, reliance on weak ties, or codified rules.
Mega-bands. 500 ± 150 people
Tribes. 1500 ± 500 people
Group trust, much like friendship, forms according to a process that imposes constraints on any social design. When we matchmake a set of random players together, we first get a low-intimacy, low-trust group of strangers. We then need to take that group through a period of social norm formation and relationship building. This process creates a rich, highly predictive social contract between individuals, which enables people to depend on one another in dynamic group activities.
Storming: The group attempts to make use of disparate norms for interacting with one another. This dissonance causes conflict. This process is very similar to the reciprocation loops that occur in friendship formation
Norming: The group negotiates common norms that this group will operate by
Performing: The group is able to perform higher-dependency tasks by leveraging their newly developers rules for interacting.
This relates to Dunbar’s Layers in a few key ways:
Existing strong bonds can facilitate group norming. If there are existing friends, they’ve already negotiated a set of common norms between them. This is a foundation to build upon when deciding the group’s shared goal and social contract.
Small groups need to build fewer bonds in order to perform at high levels. You can think of the norming process as one where people negotiate some minimum level of triadic bonds between all members of the group. With smaller groups, there are fewer connections, so the process goes more quickly.
Larger groups naturally have fewer intimate bonds. When dealing with people in the outer layers of our network, we rely more strongly on official rules, rigid social norms, and other forms of bureaucracy. People stop trusting the individual and instead lean upon a system of governance. This is less efficient, in general, due to the cost of maintaining the system, but lets more people participate.
Tips for building group trust
Mentoring. People often obtain high levels of competence through interaction with a coach or mentor. Finding ways to incentivize groups to adopt lower-skill members in order to train them up will benefit both group cohesion and general communal friendliness.
Slower integration. If facilities are not provided for subdividing groups at this layer in to smaller groupings, then every new member must be inducted by introducing them into a central communication channel. This greatly reduces the chances for the new member’s retention in the group, as they must form relationships with everyone at once, rather than being adopted by a segment of the organization and then extending their relationship network outward from that solid foundation. The best groups at this size and above have clear 15-ish person clusters, which are an ideal size for integrating a new member.
Groups vary substantially in how long they last. There are two distinct types of groups worth looking for when designing your group systems:
Primary groups. Long-lasting groups of family and friends. They tend to have strong bonds and a shared sense of purpose. People usually only belong to a few primary groups corresponding with their inner Dunbar’s Layers.
Secondary groups. Temporary, task-focused groups. These can be large or small. People often belong to many secondary groups corresponding to their outer Dunbar’s layers. It is important to allow people to join (and leave) multiple secondary groups, as they need.
Large group stability
Even through group size and Dunbar’s Layers are very different concepts, they do seem to be related. Small groups are stable at around 5 people, primarily due to their heavily reliance on long-lasting family relationships. Large group sizes tend to stabilize around the 50, 150, 500, and 1500 values found in Dunbar’s Layers.
This works in two directions:
Growth: Groups below 150 tend to grow to that size.
Fission: Groups above 150 tend to fragment into sub-groups of 150 or less.
Stable friend groups
Groups at 50 and 150 find long term stability, often measured in years, by benefiting from peer pressure (norm reward and censure), without the need for complex rules and hierarchy. The stronger the sense of shared purpose, the more robust the group. There’s more research to be done here, but this seems to be the maximum group size where, due to the limits of Dunbar’s Layers, you can rely on unaugmented human nature to self-organize into stable groups.
Stable non-friend groups
Stable groups at 500 and 1500 are far rarer because they require the addition of some from of hierarchy in order to be sustainable. Usually this involves appointing a small group of 4-5 decision makers who represent other 50 to 150 member sub-groups. These decision makers represent ‘weak ties’ between groups.
Weak ties are key to the stability of 500 and 1500 player groups. They let a group of 50-150 reach out to other groups and quickly gain access to resources, opportunities, and information. Studies show having a diverse set of weak ties — particularly in a large community of uncaring strangers — increases life satisfaction.
Weak ties are not universally good for game developers.
Scope creep. The economic and political systems necessary to make very large groups function are often some of the most complex features in a game. To support weak ties in your game is to accept a certain level of scope creep.
Over emphasis on weak ties can hurt strong ties. Weak ties are also not a replacement for strong ties. Social groups involving mostly weak ties are poor at providing emotional support as well as transferring and enforcing group norms. Many critiques of strongly capitalist, technocratic or libertarian dystopias center on how a overreliance on weak ties (via large-scale trade, algorithmic replacement of reciprocation loops, and other scaleable-yet-dehumanized systems) leads to an accidental erosion of strong ties.
If anything, modern MMOs suffer from too many weak ties and not enough emphasis on building and supporting strong ties. Perhaps because MUDs and early online games were historically rich with strong bonds, MMO designers simply assumed they’d get those for free. They didn’t realize their desire to build a big game—which historically has been conflated with popularity—was antithetical to the magical social connections that made early online games attractive in the first place.
Shared goals for different group sizes
Shared goals are the single strongest predictor of group cohesion. Groups with more group pride and stronger task commitment have strong shared goals. They are most likely to perform well at high-trust tasks, and have high retention, longevity, and increased sense of member well-being.
Group pride and identity
Members with strong group pride feel strong allegiance to the group, are happy with what the group accomplishes, and promote the group identity to others. Group pride is expressed in the same fashion across different group sizes, but identity becomes more formalized as group size increases.
Weak identity. Small friend groups may not have an official identity, and many of their positive feelings come from mutual support.
Official identity. Large social groups have official identities and a strong sense of membership. When people are part of a high-performance group, they feel like they are part of something bigger than themselves, which can lead to a sense of awe.
Stereotype-based identities. At the huge impersonal scale, we see strong tribal identities and stereotypes. People build simple cartoon models of how other people should respond to interactions. Splitting people into in-group members and out-group members occurs relatively quickly, using only superficial information.
Task commitment is about shared activities that contribute to a common goal. Group pride answers, “Who are we and do I belong?” Task commitment, by contrast, answers, “What are we accomplishing by working together?”
Tactical tasks. Small secondary groups understand their purpose in terms of short term tactical tasks. This could be completing a small project or finishing an ad-hoc raid together. Small primary groups are usually focused on supporting one another.
Trying to sustain the group. Large social groups are focused on bigger topics like long-term survival or sustaining a community that upholds shared beliefs. Group vs group superiority is an interesting task at this scale, especially for groups composed largely of young men.
Part of an ecosystem. Huge impersonal groups are brought together by convenience. They share a common set of codified practices involving trade, language, and practices that help their smaller friend groups accomplish desired goals. The task commitment present at this level usually involves maintenance of support systems such as political or economic structures.
Tips for increasing shared goals
Share goals, not just shared rewards. Many game designers assume that if there is a shared reward, people will naturally align their activities. This might work if humans were hyper-rational, profit-maximizing automatons, but they are not. Instead, players benefit from clearly-stated goals and examples of how they might work together.
Public and private spaces. Large social groups are composed of sub-groups that require private space to reinforce vision and social norms as well as create opportunities for group bonding. They also need public space to display and reinforce the group’s overall identity.
Group vs. group content. Conflict with other groups is a common method of providing a shared purpose. Meaningful rivalries can play out over the course of months or years. Games with PvP content can create very rich social histories if they can operate at this scale.
Positive goals involving growth and support. Though it is easy to rely on competition in order to give your group a purpose, history is rich with high-longevity groups, usually in the form of religious communities, that exist to preserve a positive way of life. Consider how your game can be a positive refuge from the broader world. Many players will find this to be a worthy goal to dedicated their time toward.
Roles for different group sizes
Every group needs to agree on roles within society. These are composed of appropriate division of labor and division of resources.
Division of labor
Specialization increases with group size.
Overlapping roles. In small friend groups, there’s substantial overlap in roles, with a single individual performing many different activities on an as-needed basis. Cross-training and a lack of specialization is quite common. On high-trust tasks, there’s heavy interdependency and the loss of any individual is sorely felt by everyone.
Specialization. In large social groups, we start to see specialization where individuals take on specific roles. Secondary groups focused on specialized tasks are common and people belong to multiple of them. An individual may train in several roles and perform one role for each secondary group.
Jobs become identities. Huge impersonal groups see the emergence of jobs and classes. A person has one dominant job they do in a hyper-specialized economy, which becomes their formal identity within a broader, rule-based society. They are a crafter, or a teacher, or a doctor, and nothing else.
Division of resources
Economic complexity increases with group size.
Communal sharing. With small friend groups, resources are often communal in nature with substantial gifting and untracked sharing. Social currency and interpersonal trust are more important to transactions than currency.
Value-based barter. In large social groups more formalized 1-to-1 trade in the form of barter appears. There may be a local currency and individuals keep close track that each trade is of equitable value.
Complex economic networks. In huge impersonal groups, both labor and resources are traded within an economic network with markets and auctions. Trade is strongly depersonalized, with every interaction based off a standardized currency. This network is heavily dependent on weak ties and super-connectors—people who can maintain more than 150 meaningful connections—play an outsized role in keeping various sub-groups together.
Status relationships for different group sizes
Status and hierarchy start out relatively undefined in smaller groups and grow in complexity with group size.
Leaders become more important and less personal in larger groups.
Context-specific leader. Small friend groups often end up electing a de facto leader, consciously or subconsciously, who is best-suited for the task at hand. Groups this size should be encouraged to designate a leader/organizer who can help keep the members focused on their shared purpose.
Cult of personality. Large social groups require leaders. Synchronization of activity becomes immensely difficult without a central authority wrangling the various sub-groups to move in an aligned direction. At this size, leadership is largely a “cult of personality,” driven by personal relationships instead of institutional power.
Symbolic leaders. Huge impersonal groups use ceremonial leadership, where the leader is a concrete, personified symbol of shared purpose, allegiance, and/or cultural values. There are a variety of related techniques including the use of celebrity, figurehead leaders, religious characters, or heroic individuals. These establish a type of group known as a reference group, which individuals look to when determining which social norms to emulate.
Hierarchy becomes increasingly necessary as group size increases.
Fluid. Small friend group organization is fluid and often depends on the best person for the task at hand stepping up.
Activity related sub-groups. Large social groups show visible hierarchy composed of a few primary groups and a number of task-focused secondary groups. We begin to see multiple 5 and 15 person groups operating inside groups of this larger layer.
Complex hierarchies: Huge impersonal groups have complex official hierarchies. Groups need official political and economic relationships in order to function.
Tips for supporting status
Tools for status signaling. The ability to signal hierarchy and status help larger organizations function. Titles, karma points, and visual flare are all systems that allow status to be earned and displayed.
Official reputation tracking. For huge impersonal groups, we see the emergence of strong anonymity, reputation starts to be more important than actual skill competence and parity. At the 500 person layer, freeloaders and bad actors can more easily slip through the cracks, so official means of keeping tabs on someone’s reputation benefits group cohesion.
Social norm formation at different group sizes
Norm formation in social groups involves how a group determines the rules they operate by and how they communicate those rules.
Rule formation becomes increasingly formalized as group size increases.
Personally negotiated norms. Small friend groups negotiate rules on a one-to-one basis, usually through small group discussions. Often behavior is determined on a case-by-case basis depending on the person and the context.
Key decision makers. Large social groups follow the behavior of high-status individuals or leaders. One or more smaller, high-status groups make decision through consensus-building and then then share those decisions with lower-status individuals. In more equitable groups, simple voting systems appear.
Public rules. Huge impersonal groups use official legislative systems for setting or revising laws. They usually have formalized community feedback mechanisms. At this stage we see strong rule of law. In more organized groups, explicit rules become prevalent. By requiring that people work in a very specific, codified fashion, you remove uncertainty and increase the group’s ability to function. You’ve replaced the slow process of negotiating social norms, through in-person reciprocation loops, with rules that simply tell you how you should act. However, this comes with substantial downsides. These rules are inherently less flexible. They need to be written up, conveyed, and enforced. If the situation for such a group changes slightly, the existing rules may actually reduce efficiency. And there’s no real trust—all parties execute on a pattern and hope it works.
Communication shifts from reciprocation loops to broadcast as group size increases.
Personal conversation. Small friend group members will communicate frequently and in depth with all other members of the group. This communication is likely to be equally spread between all members and structured as peer communication.
Tiered communication channels. Large social groups have multiple tiered communication channels. There generally needs to be an open, shared channel; a one-way broadcasting channel from leadership; and a number of sub-group channels for specific primary and secondary groups.
Broadcast communication. In huge impersonal groups, personal communication simply cannot reach all the people in the many sub-groups, so these groups must use broadcast technologies to send one message to many people cheaply. This, in turn, enables propaganda, where various parties use broadcast media to push unquestioned messages that promote new truths. There’s no consent loop for someone to provide a contradictory response. This is useful for spreading new social norms about how one should behave or for emphasizing group bonds, but cartoon symbols of complex systems end up being easier to spread than deep understanding.
Conflicts and sanctions at different group sizes
What happens when norms are violated?
Small friend group
Personal disagreements. Small groups are constantly negotiating norms. Norm violations are typically confronted in small group conversation, one-on-one or with the whole group, and are essentially arguments.
Withdrawal from the group. In extreme cases, members of small groups will stop talking to someone with whom they have a personal disagreement, or the group will distance themselves from an individual by lowering the frequency of interaction with them.
Large social group
Cliques and bullies. Groups this size can form into abusive groups of bullies. Designing for groups this size means added community management. Groups of this size rely on vision-based leadership, which can allow hate groups and other fear-based organizations to fester.
Huge impersonal group
Demonizing outgroups. It is common for very large groups to explicitly label those who are enemies of the tribe. This is less about attacking the outgroup and more about focusing the larger group on a larger shared goal. The downside to this is it usually relies on fear, which short-circuits more thoughtful and constructive group coordination patterns.
Law enforcement. Tracking down those who break the laws and determining the best way to change their behavior is a feature of very large groups.
Economic scams. Groups emerge that profit from preying on people who want to get an economic edge. Various black market scams start to be common, like account and credit card theft. Outside groups target the community.
Organized griefing. Though individual griefers exist in smaller groups, within a larger population, a griefing tribe can satisfy all of an individual’s social needs. Griefing becomes the social norm for such people and there’s no leverage at any point in the network to deprogram a griefer. This can lead to all-out wars, where one group attempts to destroy, alienate, or otherwise expel a rival group.
Account manipulation. With a large number of strangers, it’s hard to track who is coming or going. In online games, people create extra accounts and use them for botting, muling, multi-boxing, and other techniques that are otherwise easily trackable in more intimate settings.
Internal corruption. As with nations, if a game becomes big enough, it is easy for moderators and community management to surreptitiously misuse their powers.
Game Design Insights
Considering the constraints imposed by friendships, Dunbar’s Layers, and social groups, it is worth exploring game design that is centered around natural human social scales. Human-scale design is social design that targets the 5, 15, 50 and 150 person egocentric networks and associated groups. It explicitly avoids player systems involving 500 or more players.
If you can build a human-scale game that enables a player to spend quality time with good friends, you’ll likely improve the quality of their life. While if you break these hard limits, you actively damage your game’s social systems. These social psychology models should do more than just inform our evaluation of game systems—they should be actively shaping the way we approach design.
Such an approach focuses on smaller, more intimate social design as the core of a game. It is less concerned with big numbers and infinitely scalable systems, and more interested in fostering trust and connection between players. This perspective led us to some fundamental insights concerning how we approach online game design.
Don’t build a big world first
A common pattern when designing an MMO is:
First, imagine a big world
Then, figure out what to fill it with
Finally, create all the systems necessary to support all the stuff you’ve dreamed up
As a result, the final systems are often surprisingly complex. You’ve jumped directly into designing systems that need to handle the many issues associated with 500+ groups (i.e., your player population). Immediately, you are faced with the key problem that your world is just a large, empty area where a player sporadically meets strangers they don’t trust. As conflict inevitably arises from these low-trust interactions, the dev team toils to add a vast amount of bureaucracy to manage the poor player experience. It can feel like patching unending leaks in a poorly-placed dam.
In the best cases, like EVE Online, players create their own systems of crude governance to shore up the faulty social design. But for the majority of games, we see outcomes like The Sims Online, where mob-style groups grief new players and chase them from the game.
From a social design perspective, this process sets the team up with the hardest possible design challenges, essentially creating a lot of extra problems that then need to be solved. Focusing on designing for human-scale suggests a different approach:
Define social activities. First, imagine activities/content/context for players to enjoy together. For example, you might prototype cooperative raid mechanics for a PvE MMO.
Map out group sizes and trust level. Then, figure out what group sizes and levels of friendship best fit those activities. High-trust activities should be reserved for small groups of close friends in the 5, 15 and 50 layers. Low-trust activities can work for groups up to 150 in size, but not beyond. In fact, explicitly remove activities that involve more than 150 people. For our MMO example, you take your raid prototype and map out variations of the raid that are suited for high-trust small groups, low-trust small groups, and low-trust large groups.
Build appropriate social support systems. Build systems that support the right activity for the right group size. With the MMO, you realize that your high-trust small group encounter needs high-bandwidth communication channels to execute, so you add voice chat or rich emotes—two possible communication channels that support and enable group performance.
Scale the activity based off quality and quantity of friends available. Finally, organize the activities/content so that player groups can organically scale up and down. In the MMO example, a single player might be present and you’ll want to serve them up low-dependency, small-group content. But if a stranger appears, consider how the game might switch to a low-trust activity with parallel play? If several friends appear, how would the activity allow them to opt-in to a higher-trust (and higher-reward!) challenge?
This approach has the advantage of more closely mapping to how humans have grouped historically: in nested layers of families, tribes, villages, etc. Sticking closer to the natural shape of social grouping will make your group activities feel more familiar and facilitate social bonding. It will also allow you to apply lessons and best practices from psychology and anthropology more directly.
Social design drives retention and engagement
When game designers think of retention, we often first consider User Experience (UX). Using the logic of UX, if a developer builds a complicated core interaction that is difficult for a player to understand, most players will churn out early on. Such games should have poor early retention and struggle with new player acquisition.
However, the game industry has many counterexamples. Dwarf Fortress, Go Pets, and Dofus are three games renowned for their poor user experiences. They have weak tutorials, byzantine gameplay loops, and a general lack of traditional first-time user experience polish. By all traditional UX values, they should be failures, yet they are not.
While these games have poor UX, they also have strong social design. For example, Dofus is a game that is specifically popular in France. Its developers tried to expand its reach to other countries with limited success, for many of the aforementioned reasons.
What made France special for Dofus?
Cultural event. When Dofus first launched, French-language MMOs were rare and early adopters were blown away by the novel experience.
Basic virality failed. Players actively proselytized the game to friends, but their friends weren’t able to play as the game was too difficult to learn.
High-cost transmission to close friends. So players went over to their friends’ houses, helped them install the game, and spent hours teaching them, in person, the nuances of how to play.
This was not intentional, but the result was that Dofus ended up being played predominantly by friends, many of whom were already part of each others’ 50, 15, and 5 person layers. This allowed players to build groups stocked up with high-trust compatriots and overcome the high-trust activities in the game. Succeeding at those challenging activities in groups of trusted friends gave the game incredibly high engagement.
A virtuous cycle occurs where strongly-bonded friends make a game their homebase—a safe, intimate space for acting out their friendship. In turn, those players recruit more of their friend networks into the game.
We’ve observed a similar process in other poor-UX, high-retention game examples. To be clear—poor UX is not the root driver for these games’ avid, high-trust communities. Instead, it is one of many pragmatic reasons for players to bring the inner circles of their friend networks into a game.
The reverse of the same basic process that drove the success of Dofus highlights problems with early Facebook-style virality. Such “social network games” would obsessively incentivize players to send out invites to as many people as possible. Two results occurred:
Mismatched reciprocation loop costs. Incentivized by these games, players made a low-cost overture to a friend or associate (an invitation) that required a high-cost response (registering for, and playing, the game). This is a huge no-no when acting out reciprocation loops; it actively damages a relationship by suggesting you are ready to extract value from your friend instead of building toward future shared need. That is to say, it annoys your friends and makes them question the value your relationship.
Dilution of community trust. Second, it brings low-trust people into the game. Because Facebook didn’t care about Dunbar’s Layers, especially in the early days of the social gaming boom, many users had social graphs with hundreds of “friends,” many of whom were no better than strangers. The low-cost overture to join was little better than a random spam ad and brought many of those random players into the game, diluting the level of trust for the community already inside the game. In one fell swoop, this greedy practice hurt retention, engagement, and future growth.
All of these examples highlight the basic truth that social design is deeply powerful, but is often not a first-order consideration for designers.
Use proper terminology
A very common confusion that came up many times during our discussions was the difference between friends, Dunbar’s Layers, group size, and concurrency (the number of players simultaneously logged in). These are all four distinctly different concepts, yet it is common for social designers to use them interchangeably.
Much of this is the fault of our existing terminology. When we talk about multiplayer games, a common shorthand is to say, “It’s a 16-player game.” We all know that this means there are 16 concurrent players in a match or room, but we often erroneously assume that this also means they are all friends and/or that they are all part of the same social group.
Both of these errors are a naive misunderstanding.
Concurrent players can be spread across multiple types of social groups. Some of them might be members of groups that are antagonistic to other sub-groups in the game. Some are members of multiple groups. Some form small sub-groups while others form large sub-groups.
They can have a mix of friendship bonds. Some of them may be friends. Most are likely total strangers.
In general, having 16 people online together says almost nothing about whether or not they are in a group, or what the strength of their relationships might be. It is tempting to fall back on old, inexact language, but your game will suffer. Instead, teach your design teams about friendship formation, constraints on types of friendship, trade-offs involved at different groups sizes, and the logistics of social play.
Use Dunbar’s Layers to determine the level of collaboration your audience will support
The structure of Dunbar’s Layers gives us insight into how many friends of a given trust level you can expect a player to have online at any particular time. There are logistical implications for matchmaking, events, and more.
At the most basic level, the logistics of Dunbar’s Layers help you predict the outcome of the following example:
You design a high-trust activity that requires 100 people.
But we know from Dunbar’s Layers that any human being will only have a maximum of 15 people in their life that have this particular level of trust.
You’ve created a logistics mismatch that will results in inevitable failure. And you didn’t even have to build the game, launch it into the market, and watch it fail. You just saved your team millions of dollars and years of their life!
However, we can gain more detailed insights. Here’s how you calculate the exact portion of a player’s friend graph you can actually address with your game. First, you’ll need a few pieces of information:
Share of social time
Distribution of friends
Share of Social Time
Share of Social Time is the percentage of a player’s total time spent socializing that is spent inside your game. This corresponds roughly to the percentage of a player’s social graph that is active in the game. If a player spends 50% of their social time in a game, we’d expect roughly 50% of their friend network is also in the game.
There are a couple ways of calculating this. Conservatively, we know from time-usage studies that the average American has approximately 5 hours of leisure time per day . From this perspective, Share of Social Time equals Hours per Day Spent In Game / 5 hours.
However, less conservatively, we know that people tend to spend approximately 0.65 hours per day actually socializing. This is likely an underestimate since the time-usage studies don’t measure time spent socializing at work. Nor do they consider time spent in playing games  as socializing.
For the following calculations, we’ll use the conservative definition of Share of Social Time. For comparison, the heaviest players of Fortnite, around 8% of the player population, spend 3+ hours playing per day. That’s roughly 60% (or more) of an average American’s total leisure time.
Concurrency ratio is the ratio of monthly active players (MAU) to those currently online. Since synchronous activities require people to be present, it does us no good if you have friends in a game, but they aren’t actually playing.
A highly-social MMO will have a concurrency ratio of 10:1, so for every 10 MAU you’ll have 1 of those players online. An international phenomenon like Fortnite enjoys a 20:1 ratio, while many web-games are as low as 150:1 or 250:1.
Distribution of friends
Dunbar’s Layers suggests that our relationships map onto a very specific frequency distribution of friends.
Chart 1: Percentage of friend network layers present in the game
This distribution holds true only if we make several assumptions:
Long-term engagement. First, our game is a long-term activity which has been going on long enough that inner layers like intimate friends or best friends have grown in the game or have integrated pre-existing, external friendships. If a game is new or people have been playing for less than 200 total hours , you’ll see this distribution shift towards casual friends and strangers.
Sufficiently large cohort. The total population of monthly active players is at least 1500.
Support for all layers. If your game doesn’t have all appropriate social mechanisms for any given layer—such as the need at the 5 person layer for private locations/communication to facilitate safe disclosure—that layer will be less represented.
We can use Share of Social Time, Concurrency, and distribution of friends to calculate some useful information about our game.
Let’s say you have a highly engaging MMO:
Share of Social Time: 50%
Currency ratio: 10:1
How many friends will be in the player’s friend list? Given the standard distribution of friends, 50% of that player’s social network will be present in your game. With a total of 150 friends that means there will be 75 friends playing the game.
How many friends will be online right now? Of those 75 friends in the game, due to the concurrency ratio, only 10% (7.5 friends) will be on at any point in time, on average.
What type of friends will be online right now? Using the distribution of friends in various layers from the chart above and multiplying them by the total friends online, we can expect the following distribution of friends:
Casual friends: 5
Good friends: 1.8
Best friends: 0.5
Intimate friends: 0.3
This sort of calculation puts much harder constraints on the types of activities that we can build into our game. Note that this is a best-case scenario. A highly-social MMO with great concurrency, and a player with a fully-engaged friend network. In this best-case situation you are lucky to get a single good friend playing alongside you. You will however get a few casual friends.
This suggests that the core activity of even highly-social games with long-term, highly-invested players should predominantly be target low-to-moderate trust activities involving 5-7 players.
What does this distribution look like at different cohort sizes? Using the same logic, you can see what friend distributions would look like at various fixed populations of active players.
Chart 2: Max and Average number of friends an individual will have for various cohort sizes in a game with 50% share of social time and 10:1 concurrency.
Due to the logistics of concurrency ratios and Share of Social Time, we max out the number of friends online at around 1500 people in a cohort. Simply having bigger cohorts doesn’t improve friend concurrency.
How can we improve these numbers?
The previous calculations are just an average of the sort of friends you can expect online. By shiftings a few variables around, we can create much higher densities of friends.
Events. A timed event or a scheduled boss raid spikes the number of people online and can dramatically reduce the concurrency ratio. If you can drop the concurrency ratio to 2:1 with an event, then you have upwards of 12 friends and 4 good friends playing. This shift is one reason why events like boss raids can be high-trust events.
Asynchronous Activities. Activities that people can do when others are offline allow for more people to be involved. Some asynchronous activities can reduce the concurrency ratio to the equivalent of 1. These systems have the downside of dramatically slowing down reciprocation loops by reducing communication bandwidth, so building out a full friendship network may take longer for players.
Recruitment. Given the low engagement of the innermost friendship layers due to simple logistics, it is unwise to rely on close friends naively playing the game together. Invest in systems that actively encourage players to play with loved ones. Give them tools for scheduling these activities.
Relationship design as systems design
By translating fuzzy social psychology concepts into more mechanical concepts, we can start treating social design as a form of systems design. (Some may find the term ‘social systems design’ more palatable than ‘social game design’ after dealing with the horrors of Facebook.)
In particular, social design benefits from using the internal economy perspective, where relationships are modeled as resources and transformations on those resources.
Each relationship between two individuals is a pool. A pool is a container that accumulates resource tokens.
Successfully completed reciprocation loops is a source that produces a resource called social capital that accumulates in each relationship pool.
Rejected or unequal reciprocation loops are a sink that depletes social capital. As does distance and lack of contact over time.
Dunbar’s Layers act as a cap on the maximum number of each level of relationship you might have. When a relationship pool fills up in one of the outer layers, it may transform into a new pool in one of the inner layers. However if the inner layers are full, one must give. If any of the layers are empty, the player seeks actions that fill them.
This paints the process as rather cold and transactional. In practice, this type of design drives intense emotions. Losses of social capital yield strong negative emotions, while gains generate positive emotions. Rate of lose or gain will dramatically intensify the emotional response. If your goal is to make players laugh, cry, or otherwise experience the peak of what it means to be human, build strong social systems.
Minimize designs that require huge impersonal groups
When we develop a game that involves group sizes of 500 and 1500 people, we’ve created populations beyond the human brain’s ability to understand other people through personal relationships. Our players know nothing about most other individuals, as they are incapable of building a large-enough social network to understand the whole. Instead, they must rely heavily on rules and heuristics to govern their interactions, and we, as game designers, are on on the hook to provide those structures.
By simply upping the size of our community, we’ve introduced an immense design challenge. We now need to build systems to manage crime, corruption, economic complexity, classism, racism, and more. Suddenly, our games exhibit most of the ills of modern society and the burden is fully upon us to solve them. If we don’t conscientiously address these issues, our community collapses into a hellish online dystopia.
If you care about maximizing social impact while minimizing scope:
Consider building communities of 50–150 players. This will maximally leverage strong bonds for retention and engagement.
Use instancing to ensure that your game can support a massive population even though each community is self-contained. Games like Minecraft; Don’t Starve Together; old, instanced MUDs; and numerous other small community games suggest this strategy can be both financially successful and fulfill social design goals.
If you want to create larger communities, try limiting yourself to cohorts of 500–1500. There are no other systems larger than these values that are meaningful on a relationship level, and by creating larger populations, you dilute and harm existing social bonds.
When creating groups of 500–1500, leverage your instanced groups of 50 and 150. Create a few low-scope systems that allow weak ties between strongly-bonded friend groups. Trade networks and information exchange will be among the highest-value systems to invest in.
Opportunity: Serving Player Motivations
Games that thrive are almost always ones that satisfy a strong audience motivation. This is no different for social games and social features. Dunbar’s Layers, in particular, give us a structure for understanding the player’s social motivations.
The Belongingness motivation
“The belongingness hypothesis proposes two main features. First, people need constant, positive, personal interactions with other people. Second, people need to know that their bond is stable, there is mutual concern, and that this attachment will continue.”
You can think of the various relationship layers as a slots in a list. Everyone has space for about 5 intimate friends, 10 best friends, 35 goods friends and 100 casual friends. If those slots are filled with healthy, mutually-beneficial relationships, a person is reasonably happy.
However if any of those slots are empty, people have a strong desire to fill them in. When they don’t have those slots filled they tend to be unhappy, and, in response, will seek the company of others using several key strategies:
Deepen bonds with existing friends. This is done in order to fill inner layers of the friendship network.
Meet new people. This is done in order to fill outer layers.
Become a member of a group. Often belongingness will be combined with a desire for affiliation. By becoming part of a social group, it becomes substantially easier to both meet new people and quickly deepen friendships. Think of group membership as a bonding multiplier. It is easy to get caught up in group affiliation as an end, by itself, but remember that, ultimately, people join groups not for the sake of the group, but to fill gaps in their primary friend network.
The desire to form relationships waxes and wanes
Life events are predictive of gaps in a person’s friendship network. As new people show up in a person’s life, there’s less time for activities that require making new friends.
Entering a new intimate relationship or marriage. This fills an inner-layer slot. There’s also the inevitable shifting and merging of your two friend groups.
Having a child. This also fills an inner-layer slot. All that time spent in parenting groups often shifts friendships from your single friends over to other parents with kids just like you.
Getting a new job. This can fill any number of slots in several layers as you form new work relationships.
What loneliness looks like in a thinned-out network
There are also numerous events that thin out a person’s network.
Becoming unemployed. You lose work relationships.
Retiring. This is similar to becoming unemployed, but often you lose professional associations as well.
Breaking-up or divorce. One of the more intense losses of an inner, highly-intimate bond. As well as a weakening of all the shared relationships (closed triadic relationships in your networks).
Moving. Shifting many high-intimacy friends into outer layers. Can break existing friendships and free up slots. Research suggests it generally doesn’t destroy intimate family bonds.
Kids moving out. When kids go off to college, most parents end up losing key members of their inner circle.
Becoming elderly. There’s a slow erosion of existing friend networks as people move or die. Elderly are also are less mobile and thus struggle to meet new people.
In particular, there seem to be three major periods in which loneliness spikes: Late 20s, mid 50s and late 80s. During these times one study reported as many as 75% of people report being lonely. These values hold across genders. Providing these individuals with tools for building healthy relationships would be immensely beneficial to society.
Two social game design opportunities
All of this suggests opportunities for social game design to improve the lives of our players.
Games for friends. High-trust games should target those with free time and strong, existing friend networks. The design focus is on bringing those friends into the game.
Games that help make friends: Games that deliberately try to convert strangers into better friends should target groups that have gaps in their social network. For example, one demographic might be lonely 50-65 year old men who are seeing an erosion of their social network due to unemployment, kids moving out, and fewer opportunities to find new friends. Make a game that is the modern version of a Masonic Lodge.
Both opportunities could be served by the same game, but be sure to sort incoming players based on their needs and direct them into activities that satisfy those identified needs.
The big idea
Key discoveries in social psychology place hard limits on the types of social games we can build.
Friendship research shows meaningful in-game relationships require conditions such as proximity, similarity, reciprocity, and disclosure
Dunbar’s Layers research shows that players have hard limits on the number of meaningful relationships in their life. These friendship are organized into layers of increasing size and decreasing intimacy.
Social group research shows the need for increasingly complex support structure as group size grows
These are the physics that social designers must understand and build into their designs.
Many past designs ignored Dunbar’s Layers and naively assumed “more is better.” They ignore friendship formation and assume “it just happens.” They ignore social groups and arbitrarily mash players together.
In reality, these assumptions are actively harmful and cause the following:
Fewer in-game friendships. A flood of strangers swamp the reciprocation and proximity mechanisms that generate friends. Poor identity, persistence, reciprocity, and consent systems mean these strangers never convert into friends, so there are fewer meaningful relationships in the game.
Increased toxicity. Large groups of strangers naturally breed toxic sub-groups. Players engage in violent rejection of out-groups in order to protect their experience and intergroup conflict becomes the cultural norm. Such communities are hard to reform and poison long-term retention.
Scope creep. The additional systems necessary to manage large groups of strangers substantially increase the scope of your game.
What players need
If players have not filled all the slots in their primary friend network, they suffer. And, in response, they are intrinsically motivated to deepen their existing relationships or build relationships with new people. Striving for belongingness is one of the strongest human motivations. They will naturally seek out activities that help them make friends and belong to something bigger than themselves.
If your games help build relationships for the player in any of their inner layers, you’ll accomplish a couple key benefits:
Increase retention and engagement. Your game becomes the place where people attain their desires. Since you provide immense value, they make the game a key part of their lives.
Improve the lives of your players. They’ll experience less depression, better health, and have more robustness in the face of negative life events.
If we take all the insights gleaned from research into group psychology, examples from online game design, examination of Dunbar’s Layers and social motivation—all of it into consideration, we can arrive at several, strong best practices:
Build games for smaller cohorts. The base activities should target small, collaborative groups. Large groups of close friends are rare or, in many cases, mathematically impossible.
Cluster players into persistent, high-density cohorts. So they have repeat interactions with the same players. The more reciprocation loops that are completed, the stronger the friendships. Big, empty spaces are not a positive feature.
Encourage high-concurrency events or asynchronous activities. Logistics favor players being around to interact with their friends. Having friends playing the same game doesn’t matter if you never see them.
Aim for long-term engagement. Build a game where players are engaged for hundreds of hours, so they have enough time to build deeper friendships. It takes at least 50 hours of interactions to form a basic friendship.
Attract existing friends, if possible. Existing friends from the strongest foundation for your game community, especially when first launching your game. Put people into safe, guild-like structures and encourage them to bring in their friends.
Design for climbing the trust spectrum. When introducing strangers into your game, build low-trust activities that scale into high-trust activities. Start with parallel or single-player gameplay and allow players to opt-in to higher-dependency activities. If players start forming strong friendships in game, support them. Bring those relationships into safe places with tools for enabling consent, support, and disclosure.
As ethical game designers, we should strive towards some higher purpose beyond merely extracting money, time, and energy from our players. Building friendships and providing lonely people with human connections are goals worthy of our highest-quality work.
If you are working on a multiplayer game, ask yourself how your designs help build social capital with and among your players. If you encounter people who believe that “more is better” when it comes to building social systems, we recommend you send them this report. There’s a new wave of social game design inspired by lessons from social psychology and we are immensely excited to be part of it.
 Dunbar’s Layers. “Generally speaking, humans each have one to two special friends, five intimate friends, 15 best friends, 50 good friends, 150 “just” friends and 500 acquaintances. Our relationships form a series of expanding circles of increasing size and decreasing intensity and quality of the relationship.”
Woodward A (2017) With a Little Help from My Friends. Scientific American. Retrieved December 27, 2018, from https://www.scientificamerican.com/article/with-a-little-help-from-my-friends/
 Dunbar’s Number. “The figure of 150 seems to represent the maximum number of individuals with whom we can have a genuinely social relationship, the kind of relationship that goes with knowing who they are and how they relate to us. Putting it another way, it’s the number of people you would not feel embarrassed about joining uninvited for a drink if you happened to bump into them in a bar.”
Dunbar R (1998) Of Brains and Groups and Evolution. In Grooming, Gossip, and the Evolution of Language (pp. 80-105). Retrieved December 26, 2018, from https://books.google.com.ar/books?id=nN5DFNT-6ToC&pg=PA77&redir_esc=y
 Unequal dyadic bonds. “When analyzing self-reported relationship surveys from several experiments, we find that the vast majority of friendships are expected to be reciprocal, while in reality, only about half of them are indeed reciprocal.”
Almaatouq A, Radaelli L, Pentland A, Shmueli E (2016) Are You Your Friends’ Friend? Poor Perception of Friendship Ties Limits the Ability to Promote Behavioral Change. PLoS ONE 11(3): e0151588. http://dx.plos.org/10.1371/journal.pone.0151588
 Loneliness impacts longevity. “…individuals with adequate social relationships have a 50% greater likelihood of survival compared to those with poor or insufficient social relationships. The magnitude of this effect is comparable with quitting smoking and it exceeds many well-known risk factors for mortality (e.g., obesity, physical inactivity).”
Holt-Lunstad J, Smith T, Layton J (2010) Social Relationships and Mortality Risk: A Meta-analytic Review. PLoS Med 7(7): e1000316. https://dx.plos.org/10.1371/journal.pmed.1000316
 Friendship impacts life satisfaction. “…the results indicate that both having/meeting friends and good-quality friendship relations are important to an overall life satisfaction.”
Amati V, Meggiolaro S, Rivellini G, Zaccarin S (2018) Social relations and life satisfaction: the role of friends. Genus, 74(1), 7. https://genus.springeropen.com/articles/10.1186/s41118-018-0032-z
 Friendship reduces depression. “People who have close friends and confidants, friendly neighbors and supportive co-workers are less likely to experience sadness, loneliness, low self-esteem and problems with eating and sleeping. Indeed, a common finding from research on the correlates of life satisfaction is that subjective well-being is best predicted by the breadth and depth of one’s social connections.”
Helliwell J, Putnam R (2004) The social context of well-being. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences, 359(1449). https://royalsocietypublishing.org/doi/10.1098/rstb.2004.1522
 Social media doesn’t expand our friendship capacity. “The fact that social networks remain about the same size despite the communication opportunities provided by social media suggests that the constraints that limit face-to-face networks are not fully circumvented by online environments. Instead, it seems that online social networks remain subject to the same cognitive demands of maintaining relationships that limit offline friendships.”
Dunbar R (2016) Do online social media cut through the constraints that limit the size of offline social networks? Royal Society Open Science, 3(1). https://royalsocietypublishing.org/doi/10.1098/rsos.150292
 Intimate relationships best predict health. “…the presence of an intimate relationship (as opposed to a broader social network) [has] the greatest effect on explaining variance in depressed mood.”
Roberts S, Arrow H, Gowlett J, Lehmann J, Dunbar R (2014) Close Social Relationships: An Evolutionary Perspective. In R Dunbar, C Gamble, J Gowlett (Eds.), Lucy to Language: The Benchmark Papers (pp. 151-180). Oxford: Oxford University Press.
 How investment shapes social graph distribution. “the strength of a tie is a (probably linear) combination of the amount of time, the emotional intensity, the intimacy (mutual confiding) and the reciprocal services which characterize the tie”
Granovetter M (1973) The Strength of Weak Ties. American Journal of Sociology, 78(6), 1360-1380. Retrieved December 28, 2018 from https://www.jstor.org/stable/2776392
 Available leisure time. The average American woman spends roughly five hours per day on leisure activities (35 hours per week), while the average American man spends about 5.5 hours per day (38.5 hours per week).
Bureau of Labor Statistics, U.S. Department of Labor (2018) American Time Use Survey — 2017 Results. Press Release for the Bureau of Labor Statistics. Retrieved December 20, 2018, from https://www.bls.gov/news.release/pdf/atus.pdf
 Why are people socializing in games? “On the face of it, this may seem like a sad state of affairs. It could even be read as dystopian: people are escaping real life to be in virtual worlds. People often find community within gaming worlds, and may get a heightened sense of shared experience from competing against or teaming up with people across the world who share their interests. In some cases, these connections might even be more valuable than, say, gossiping with a neighbor.”
Kopf D (2018) Americans are socializing less and playing more games. Quartz. Retrieved December 28, 2018, from https://qz.com/1320344/americans-are-socializing-less-and-playing-more-games/
Depp C, Palmer B, Glorioso D, Daly R, Liu J, Tu X, Kim H, Tarr P, Yamada Y (2018) Serious loneliness spans the adult lifespan but there is a silver lining: Feeling alone linked to psychological and physical ills, but wisdom may be a protective factor [Press release]. Eureka Alert. Retrieved December 28, 2018, from https://www.eurekalert.org/pub_releases/2018-12/uoc–sls121218.php
Williams, D. (2007). The impact of time online: Social capital and cyberbalkanization. CyberPsychology & Behavior, 10(3), 398–406.