Ethos Before Analytics

By Chris Birke

[In this article, designer Chris Birke takes a look at research, examines what's going on in the social games space, and argues for an approach that puts creative ethos before data-driven design -- but without ignoring the power that game designers wield over players.]

A little over 10 years ago I read an article titled Behavioral Game Design written by John Hopson. Now, looking back, I see what a huge influence it's been on my game design philosophies. I have been following psychology and neuroscience ever since, always uncovering new ways to incorporate them into an ever-growing design toolbox.

Technology is moving very fast, and 10 years is very long in internet time. Wikipedia was launched 10 years ago; Facebook, World of Warcraft, and Gmail have been around since 2004. The iPhone was born in 2007, and FarmVille recently turned two. Science is moving just as quickly, and behavioral theory is now being underwritten by neuroscience, and the revelations of high resolution, real-time brain mapping (fMRI). On top of this we now design with the aid of analytics, the real-time data-mining of player behavior. We can roll out a design tweak once a day, if necessary, to maximize profits.

What are the ethical implications? As a curious proposition, "Behavioral Game Design" seemed innocent enough. Now that design toolkit verges on a sort of mind control, and the future is promising only refinement of these techniques. What are we doing to players, and what have we left behind in those innocent days of chasing "the fun"?

Personally, so long as I can make enough money to eat (and maybe have a good time), I feel obligated to design socially responsible games that benefit the lives of players, not just exploit them. I want to explore ideas of how to use these new technologies in a positive way, and to encourage those who feel the same.

I would like to share some of the neuroscience that attempts to explain how conditioning behavioral conditioning works in games, and go into how this can be used in the context of analytical game design to maximize player compulsion. Then I will go into some ideas for how to use these tools ethically, and hopefully inspire discussion in our community. But first, a brief review of behavioral conditioning.

"They're waiting for you, Gordon, in the test chamber..."

Most behaviorists don't use the words "Skinner Box." Skinner himself didn't want to be remembered as a device, preferring to call it an "operant conditioning chamber." It is a cage used to isolate the subject (usually a pigeon, or a rat) with only a button to operate and a stimulus (a light, for example) to be learned. Pressing the operant (button) releases a reward (food), but that's reliant on pressing it correctly in response to the stimulus.

It was with this that Skinner explored the nature of learning and, further, how to maximize or disrupt the compulsive behaviors of his subjects. The results, in short, showed that the schedule of rewards in response to stimulus greatly affected how animals (like you and me) responded to their training. The most compulsive behavior was not driven by "fixed ratio" rewards, where a stimulus meant a consistent prize for correct actions, but instead by a semi-random "variable ratio" schedule. Maybe you would win, or maybe not. Keep trying, just in case -- you'll figure it out eventually.

If you have been designing games at all in the past few years you ought to be familiar with this. Applying and combining the results of these studies have been proven to work. No one can deny the incredible feeling you get upon hearing the familiar "ting" (YouTube link) of a rare ring dropping off an enemy in Diablo. It's the combined reward of the long term chase for better stats with the instant gratification of a high pitched chime over the clank and groans of battle. It's rare and semi-random.

You can't argue the benefit of front-loading content onto the learning curve like in Rift (or any other MMO) either. Dishing out rewarding content more slowly in the late game not only maximizes its use, it's fitting nicely to the documented results of the most compelling reward scheduling. Just add some compelling random combat encounters to keep it fresh. Reviews (for example, Gamespot's Review) call this out as good design, because it's more fun that way, right?

Since I'm being such a depressing reductionist, let me tell you I believe there is such a thing as "fun." It's a specific brain activity within us, electric and chemical. It lives in there, and you can probably graph it with powerful magnets, sales, focus groups, or the staggering 275 million daily active users playing Zynga's games on Facebook (AppData). Even if you don't think current Facebook games are fun (and I'll suggest how that might work), someone out there does.

What is fun, anyway?

In my opinion, neuroscience is quickly extending behavioral theory as the most effective means of manipulating people (players). There are a few different theories of what's happening in the brain to create the consistent results found in behaviorism (and FarmVille), but I'll only share my favorite for sake of brevity.

If this isn't the true mechanism of fun, I'd at least like to warn you: it will be discovered soon. A early paper on the topic entitled "Predictive Reward Signal of Dopamine Neurons" is the sort of thing that makes me giddy. This research describes in detail how the behavior of a particular type of neuron in the brain specializing in the neurotransmitter dopamine works as the "reward system" to drive learning and motivation. It's fairly simple theory called "incentive salience," and the key is novelty.

All of our brains are similar. Just as the average person is born with the same sorts of cells in the fingernail cuticle on their ring finger, so, too, do we all share the same brain areas. They work to perform the same tasks in all of us (moods, facial recognition, Counter-Strike, etc.). They've specialized.

An important central structure, the ventral tegmental area (VTG), is made up of neurons that specialize in the release of the neurotransmitter dopamine. It stretches out into other brain areas, lining them and waiting for a queue to act. By releasing dopamine, this structure can intensify brain activity in those areas, acting as a sort of throttle. And what's controlling the throttle? Reward.

(Fig. 1) The reward system.

These rewards are the same sorts of delicious rewards given in Skinner's Behaviorist research, as well as other things we're wired up to like. (Social status, pleasant noises, sex, explosions, epic loot, etc.) These things trigger the signals dopamine neurons are carefully monitoring, and each expects a precise level of expected reward.

When a surprising reward occurs (or, as they say in the literature, a "salient" one), a flood of dopamine is released. That flood is like adding fuel to a fire, and the brain activity is intensified along the dopaminergic pathways.

While that's happening, new memories are being formed, too, encoding the current input from the senses as a stimulus. It's the pattern of whatever the brain was sensing and thinking while that new reward was experienced.

From now on the link is made, and whenever this pattern (this stimulus) is present, the memory of the reward is activated. This happens even before the reward itself is consumed. It's what behaviorists call conditioning.

I'll use ice cream as an example of this process. It's full of creamy sweetness (which qualifies as an excellent reward), but pretend you don't know that. Imagine you've never before in your life so much as heard of ice cream. Then take a lick.

The sugar should trigger an automatic reward signal in your brain, and your dopamine system will detect it and light you up right afterwards (Trial 1, in the figure below).

The experience of the sweetness is intensified by the heightened activity, so the memory will be vivid. The sight and smell of ice cream alone will be enough to remind you of its taste, and how good it made you feel.

The reward of ice cream is now conditioned. This is the experience of "pleasure," but I wouldn't call it fun yet. There's one last step in the process.

(Fig. 2) The dopamine release moving back in time over four trials until it reaches the furthest reliable signal of reward.

Because you've been conditioned, seeing or smelling ice cream (the stimulus) again will now activate a dopamine response even before you taste it (Trial 2). Not in the original area (those neurons closest to experiencing the taste have already begun to expect it, and aren't as surprised). Instead, new areas that weren't involved in the exact moment of the ice cream reward are being lit up.

Remember, this is now happening before you've actually gotten the ice cream. The dopamine driven brain fire is happening while you're standing IN LINE for the ice cream. The memory and stimulus is being recorded progressively further out across time (Trial 3). It's your brain encoding a prediction of the reward. Even though standing in line is not at all the same as eating ice cream, you're doing it because you're compelled by the memory. That's what we call "fun."

"But wait," you say, "standing in line sucks!" (And you're right.) It sucks for you and me because we've done it too many times, but try to imagine the first time you stood in line for ice cream. I'd bet you were giddy with expectation and could barely contain your excitement. That's the dopamine system at work. It's intensifying the experience by making the memory of the ice cream so vivid that you can almost taste it, while laying down a memory of walking up to the ice cream shop, looking over the flavor choices, and trying to decide which topping (chocolate fudge or whipped cream?)

All of these experiences and choices are new processes in your brain, and all of them are receiving an novel link to the reward of ice cream. That's keeping the dopamine flowing, and making it all feel good. As long as a reward can be predicted by a new stimulus, the dopamine system will keep causing the excitation needed to record it, and the whole experience carries a compelling feeling and expectation of the reward. It is not the reward itself, but the fresh process of learning to get it and the motivational feeling of wanting the reward.

So that's what I've come to think of: the dopamine-enhanced experience of reward while a new predictive stimulus is being mapped out. A human software interface. Although every brain is born different, some nearly immune and others very susceptible, the neurons and dopamine are always there. With a little deduction it's a tool of strong game design. Keep the dopamine flowing by making sure you provide salient rewards and just enough operant complexity to keep it all from ever being mapped out, but not too much or else you shut them down.

It explains why you're willing to keep clicking "reconnect" when the WoW server goes down, and it explains why you can rehash old mechanics so long as you're selling them to an audience who hasn't already learned them (i.e. children, new gamers.) It also establishes how splitting your rewards across many brain systems (social, aural, mechanical, visual) is such a compelling combination. It's a bit surprising that Vegas hasn't figured out how to turn gambling into more of a social game for people besides the high rollers.

(Fig.3) A momentary depression occurs when an expected reward is absent.

Once everything that can be reliably linked to ice cream has been wired up, the dopamine system goes quiet. The neurons involved have become acclimated; they no longer get excited when they sense the stimulus (Trial 4). They simply expect their reward of the ice cream in due time. With nothing new to learn, they no longer trigger excitation. Standing in line for ice cream now sucks.

Even though ice cream itself is still delicious, the line is simply work you do to get it. Things "just don't feel the same anymore." Even more than that, those dopamine neurons are still watching, and if the predicted reward doesn't show up when they expect it, the whole system flat-lines. When they fall on their face, you have no motivation to continue, and quite possibly feel pissed. It's a theory of boredom, as well as fun.

It hints at addiction too, but in doing so it resolves the difference between an "addictive" game and a drug. It has to do with the strength and means of the dopamine release. Almost all addictive drugs like methamphetamine affect dopamine, and can trigger its immediate release at tenfold normal levels. Given how this reward system operates, you can imagine how strong this conditioning is. Games only trigger the release of dopamine through normal sensory channels, and at a more healthy and sustainable rate.

By this theory, the largest dopamine rush accompanies the first few times you play. If you've ever felt a compulsion to rush home and play a game all day, you are likely getting a hint of what a strong addiction feels like. Afterwards, the combined effects of acclimation and inhibition curb further dopamine release.

Can a game lay down enough conditioning over time that begin to match the levels of reinforcement seen in drugs that immediately release dopamine? Perhaps, and we have strong profit motivation to see if this is true. It's reasonable to assume some people can be trained to compulsively play games.

Mario Collects Coins

Every arcade cabinet collected a certain number of quarters in a month. At month's end, the arcade owners counted these up, paid their bills, and sunk the rest into their futures. These were the people who decided which new arcade machines to buy.

You can imagine the effect this system had on game design, and you can think of it as a primitive form of analytics. The majority of analytics in current Facebook games are essentially the same, only one step away from counting profit. However, with the realities of digital distribution and real time metrics, this world is quickly changing.

As a refresher, analytics is the process of crunching data to inform decisions. They came to games via web developers, as the early Facebook games were not created by the traditional console studios. Web developers knew the value of analytics in a way the console world is only beginning to appreciate, and built them into the game.

Webpage "hits" and "bounce rate" translate fairly well into the world of Facebook games as "Daily Average Users" and "Monthly Average Users," and the combo stat "DAU/MAU." (It's used to measure retention.)

Beyond helping producers sound important, they provide a clear and inarguable metric of success: points on the graph! Through simple metrics like these you can see in real time how quickly your game is gaining users, keeping them, and turning a profit.

However, analytics used at the level of DAU are fundamentally limited. They show only the present, and say little or nothing as to why and how the game is creating these numbers. It's through a combination of design insight, deeper analytics and experiment that takes it to the next level.

We're putting games in the cloud now. Steam was one of the early digital distribution channels, and got into analytics right away with its hardware surveys. Facebook has more advanced demographics and built in analytic functionality (not to mention its pioneering work in redesigning our concept of privacy).

Android and iOS devices all deliver through the cloud. Google has numerous online APIs for crunching analytics, and I'm just waiting for the day websites like OkCupid start hosting apps. (It's an analytics goldmine!) The current and next generation consoles all have digital channels, and I don't see any indication of this letting up.

The more we continue to shift (or float) in that direction, so too will the analytics become easier and easier to access, and what's more, with the social graph that accompanies it all. The relationships between all the players, their in-game interactions, and the histories of players over time and across games can now be tied to a unique profile for each user. This massive well of information is where web analytics and game design begin to synthesize into something new.

Because games offer a much deeper level of user activity than websites, you can use analytics to probe much deeper into player behavior, too. Will Wright was a pioneer in using analytics for game design long before we'd dreamed of Facebook. In a fantastic 2001 interview with Game Studies he spoke of how data revealed two main types of play in the Sims: House Building vs. Relationship Building. The game was then tailored to further support these two player goals.

He went on to speak of how analytics might one day be used to customize games on a per-user basis. It's something now possible, and in real time. Rather than just gathering information on player attendance, every aspect of player behavior can now be collected as analytical data. Questions of what, when, who, and why can be asked, and a whole slew of predictive information can now be developed.

Dividing players into different demographic groups allows developers to tailor (both in content and mechanics) to particular types of players. If charging $10 for the rare Purple Penguin in Zoo Collector will not make up for your monthly budget deficit, then introduce the $10 Violet Bugatti in the nearly identical Car Collector (catering to a segment of the Middle Eastern audience with disposable income).

Not all demographics are of equal worth. By mastering specific content for a number of smaller demographics you can expand into niches beyond the mainstream, or divide the mainstream up into more manageable developmental targets. Not all players are of equal worth, either.

It's very valuable to identify your early adopter players and the opinion-generating players who act as social hubs. These players are the agents of virality, and your means to success. Identify them and let them know how special they are with rewards and privilege. Tell them early about your next game, and encourage them to move on to it. The rest will follow.

Don't forget this can all be automated, too. Games on a server can be patched twice a day, if needed, and there's no reason every user will see the same version of your game. Not only can you give your most popular users the "Special Version," you can also release "Design Version A" to fifty percent of your users and "Design Version B" to the rest.

Watch the analytics to see which version does better, then shift over the entire game to that design. Wash, rinse, repeat. This is known as "A/B testing," and you can use it to explore almost every aspect of design. Just be careful not to test various costs on the same digital content, as people might compare notes on the forums and get suspicious.

I see no reason to be so blatant, though. Rather than being reactive -- staring at your feet while trying to get ahead -- find success through forward-looking design. As outlined earlier, we now have a strong theoretical framework for interfacing with players and developing a compulsive play response through behavioral theory and neurology. A/B testing, demographics, social graphs, and an endless stream of willing players is the perfect laboratory setup for perfecting such designs.

I would recommend experimenting on subsets of users to see how well they condition to certain mechanics such that the timing and distribution of rewards can be optimized. Hire or train statisticians. Users should be tracked according to their sensitivity, and targeted to maintain engagement.

Rather than rehashing the same designs over and over, predict the effectiveness of certain designs on specific market segments and transition them from an old design to a new, more complex one the moment they are ready. This can be verified as increased retention via the MAU/DAU ratio. Meanwhile, the old design can be reapplied to a fresher audience segment.

I'd recommend that if I was a tool, that is, and unfortunately it's already happening.

Who wants to be a thief?

At some point we're getting carried away. To review, we have a system funded by the success of a theory for conditioning behavior with the neurological science and imaging to refine it, and a massive social laboratory performing millions of experiments daily to get it right. That's quite a beast to be reckoned with. It most effectively targets players who've never seen a game before, children growing up online, unemployed people, and people wasting time at work. What do the games provide in return? "Fun"?

To be sure, this for-profit laboratory existed long before video game analytics. We've been manipulating each other since the beginning of time, and we are built to accept the thoughts of others (it's called communication.) Advertising long precedes gaming in refining these methods, and before that propaganda and rhetoric did nearly as well. Even writing these words is in a way of manipulating you into thinking them.

Internet gaming, neurobiology, and analytics just provide great new ways to control behavior and generate profit. It's hardly limited to games, either.

Jesse Schell projected an alarming version of the future in his "Design Outside the Box" talk at DICE, where all commerce was motivated with a point-tracking "gamification" layer, but that's now well underway too.

It's not the Buy-Ten-Get-One-Free coffee cards that worry me, but the "Rewards Program" available for every single credit card on the market. Curious name, "Rewards Program." It exists because it works.

Is that what games are now becoming, a sort of credit scam? Without narrative, without social context, without political stance, and without an opportunity for creative expression, are we just dividing our players into little boxes to be farmed?

Are we, as designers, just clicking the same square tiles over and over waiting for the coins to pop out, so we can click them too? Are you bored, or do you still have a few more years of empty clicking in you? Games that provide more value are worth more; people just need to know.

Educate consumers about the system, and show them why your games are worth paying for. Get them to shun other games. Turn games exploiting simple reward mechanics into the McDonald's food of digital entertainment, while standing up for games that deliver something worthwhile. Create games with story and create games with art. Mechanics and a theoretical understanding of fun are wonderful tools for expressing a message in a way more powerful than print, music, and film. That's ethos, and you can generate profit with it.

The future of the medium is growing this alternative. Though developers often scoff at the idea of "games as art," it is unquestionably coming up more frequently in discussions. The Smithsonian plans an exhibition of games in 2012, and the National Endowment for the Arts is now funding games that have artistic merit.

What does it mean? I rarely encounter a developer with the humanities background to engage the question, so I'll take a chance at translating it to something engineering-oriented: art creates culture. It is not the communication and manipulation of minds on the individual level (as described in this article), but a formation of shared opinions and ethics across masses of people.

Often, art creates whole subcultures of devotion, and in doing so it engages people at a level of behavioral conditioning far more advanced and comprehensive than the simple designs of slot machines and Facebook games. It's already begun with the celebration of indie game developers, the microcosm of chiptunes, and people's unwaveringly fond devotion to Final Fantasy VI and VII characters. The virtues of Link are present in a generation, whereas the ethics of CityVille will never be.

Behavioral manipulation can be used positively. A program at Yale university called Play 2 Prevent is exploring the use of games as a tool for increasing awareness of HIV in teen sexual activity. Perhaps someone will fund games for the training and rehabilitation of prisoners, as they ought to be more educational and passively reformative than cable TV. With each day passing methods such as these are becoming accessible to all developers. Articles by researchers (like this recent one by Ben Lewis Evans) are now routinely appearing on Gamasutra.

Analytics are your friend, too. You can do experiments and optimize your games in the same way as mega-publishers; just do it in a way that gives your users something that actually benefits their life. Consumers love analytics, so long as it's for their benefit. Just put the data together in a friendly package and give it back. Share out this data with other developers (maybe we need a, and collectively out compete the companies who hoard data. Design creatively, and build mechanics off social graph data to deepen interactions with real people.

Numerous services are already doing this (from Xbox Leaderboards to the OkTrends blog) and people consciously make it viral. Push your existing concepts forward with analytics, too. If data reveals a peak number of players at five pm, schedule in game events for that time. Systems that adjust the game in real-time, like the AI Director from Left 4 Dead, can be driven with analytic data too. Embrace the concept of the cloud and use this data to keep your game from becoming static.

Love before profit.


The zen and flow of play can be beautiful and life-expanding, or it can drive people into the rut of a junkie for the profit of someone they will never know. As I've learned these things, I can't go a day without seeing them in my life. I feel it when I stop playing games because I just can't finish the unskippable tutorials. I get angry when I read poorly written textbooks on topics I want to learn, but can't, as my will is needlessly sapped with boredom.

My heart goes out to the pain of kids trying to finish the dry repetition of their math work, and it goes black when I see the finely crafted advertisements for unhealthy things tagged onto the finer moments in art. I think of the millions in talent spent to make it happen. With all the zombies pulling slots in Vegas, all the hipsters swiping down on their mobiles in hopes of a new Facebook update, and all the worn paths paced by desperate animals in the zoo, I don't want to make another game like that. But I probably will.

Design your ethics into how games will interact with players. Sometimes it's okay to make something fun and compelling. Other times you'll be forced to make concessions. I've done some pretty shameful things in development. I've compromised on principles of violence against women, I've modeled munitions for the army, and I've studied very hard at how to make people keep doing things compulsively when they otherwise wouldn't.

Don't give up. If you got into game design because you love games, then fight to show it (and you will have to fight). Things are changing very quickly, and the purpose of games is created through you.

Return to the full version of this article
Copyright © UBM Tech, All rights reserved