Using Games to Tap Collective Intelligence

I’ve been mulling over an idea for several months now. It goes something like this:

Nowadays, everyone is talking about the broad potential applications of video games. Combating obesity. Managing chronic disease. General education. Employee training. Military preparedness and recruiting. The list seems endless. But one unique and important aspect of games has yet to be tapped: I believe they can effectively aggregate individual players’ actions into a form of collective intelligence.

Basis of the Idea: Wisdom of Crowds

First, a bit of background. Few things have become conventional wisdom so quickly as the Wisdom of Crowds, an idea that has captivated savvy businesses seeking to improve their internal prediction and decision-making processes. There’s great reason to believe the hype — just ask HP, which has used idea markets to generate more accurate sales predictions, or read about the Iowa Electronic Markets, which have predicted presidential election outcomes more accurately than the polls in 75% of cases.

However, despite a lasting surge in media, business, and academic interest, proven mechanisms via which to harness the wisdom of crowds remain in short supply. Idea markets have existed for many years, as have the “opinion aggregation” systems in websites (i.e. the user-generated product rankings found in Amazon.com). The chief obstacle is and always has been: how to properly incentivize the participants in a system, such that they generate meaningful, unbiased input.

Idea markets use real or virtual currency to give participants a vested interest in the outcome of their predictions. Opinion aggregation sites rely on social and cultural forces to encourage useful behavior. Unfortunately, anything other than markets based on real currency investments can be manipulated by those who wish to bias the system for personal gain (or perverse pleasure). Amazon, for example, has always struggled in the fight against “shill” product rankings. And of course, even real currency-based markets can be manipulated under certain circumstances, if the incentives are great enough.

Games are Wonderful Incentive Systems

There is, however, one well-known mechanism that does an amazing job of incentivizing people to think seriously and passionately about a given set of problems. A mechanism that compels people to meaningfully compete, against other people or against themselves, for no monetary benefit whatsoever. That’s right — video games.

For many years now, developers have been creating games that revolve around real-world problems such as resource development, political maneuvering, etc. One of the most famous of these is called SimCity; in it, players are taught to grapple with zoning issues, tax rates, etc. What if games that encouraged people to solve real-world problems (as a means of accomplishing larger objectives) were developed in tandem with corporate or government sponsors? Not “business games”, but commercially-viable, entertaining games that consumers might not even recognize as out of the ordinary?

Give Me An Example

Imagine a SimCity-esq game in which the player is given the financial reins to a region. The game could be set in a real location (i.e. California), incorporate real world constraints (i.e you can’t indulge in deficit spending forever), and could dynamically import the latest available real-world regional data via the Internet (i.e. demographic figures, current spending levels, etc). That way, when players begin a new game, they are immersed in a situation that closely resembles whatever situation California’s politicians are currently grappling with. But here’s the catch: once players get out of the tutorial phase, the game can begin recording their decisions and transmitting them to a central database, where they are aggregated into a form of “collective vote” on what actions to take (i.e. raise the sales tax or lower the sales tax). If the Wisdom of Crowds is correct, the collective choices of 100,000 game players in California (which would include knowledgeable people as well as many less-knowledgeable people) may very well be better than the choices of 1,000 Californian policy experts.

Sound like something that governments would never even consider? The French government recently announed that it would be releasing a video game called Cyber Budget that enables French citizens to play finance minister. Cyber Budget appears to be purely educational (and promotional?) in nature, but it’s certainly a step in the right direction.

There’s Gotta Be A Catch

Unfortunately, there’s one fundamental problem with this particular example: a video game player can only be as good as a simulation permits her to be. A game has to reward “correct” choices with positive feedback and “incorrect choices” with negative feedback. But, if governments could create a perfect simulation (and thus always reward the “correct” choices in any given situation) they wouldn’t need a video game to aggregate collective intelligence for them, now would they?

Put another way: games are incredibly effective educational tools, which makes them very useful in this context… but if you aren’t careful, you may end up making a game that simply “teaches” people how to win your imperfect simulation, rather than how to solve real problems! This doesn’t mean the government game absolutely can’t work for collective intelligence, but it does present a very serious design challenge.

So Was All This Just A Tease??

No, this was not a big tease. I just wanted to illustrate a potential problem with this idea early on, since it’s easy to be blinded by the possibilities, otherwise!

So how do we sidestep this pitfall? One solution is to incorporate real-world data into the game’s feedback mechanisms. If the game uses real-world, real-time data to judge the merits of a player’s actions, it should be less susceptible to simulation problems.

Consider the following example. Accurate weather prediction is extremely important to society, but the vast majority of people would never be interested in voluntary meteorology. However, weather prediction might fit nicely within the context of a turn-based game that plays out over a long period of time — something vaguely like Roller Coaster Tycoon (in which players operate a virtual theme park.) The mechanism would look something like this:

• Players are made to understand that spending heavily on advertising before a rainy day is not a good business strategy; after all, if it’s raining, customers are sure to stay away no matter what. On the other hand, spending heavily before a beautiful, sunny day greatly amplifies an already good turnout to your theme park.
• Players are given real-time weather data on their actual, physical region (in which the game is also set). They are given access to simplified meteorological tools. In game tutorials teach them how to use all this, of course.
• Players then base their advertising decisions on how much cash they have and what they think the weather is likely to be the next day. They may not realize it, but they’re implicitely “voting” for sunny weather when they spend more on advertising, and voting for terrible weather when the drop spending to zero. These “votes” are aggregated via the Internet.
• Since the game is turn-based (i.e. you take actions today and realize their full impact over the next several days), the game can take into account actual changes in the weather when giving feedback to the game players. In this way, the simulation is kept “pure” and players kept “honest.”

It’s possible that a turn-based convention isn’t necessary. Perhaps the majority of weather data in the game could be historical — i.e. players are exposed to true weather data from the past, and thus, the game knows exactly what should happen “a few days later.” However, every once in a while, the game is remotely triggered (by the developers) to expose current weather data to all players. In this situation, the game would use meteorologists’ current best guess about tomorrow’s weather in order to provide feedback to the game players. But meanwhile, the players’ collective votes could be used for predictive purposes, without too much risk of “impure simulation spoiling.” In this example, you’d get useful data on rare occassions (as opposed to every single day), but the tradeoff opens up many game design options.

Also: if you don’t like the theme park thing, weather prediction could be incorporated into a military-centric game, too. Or something else.

Economists Love This One

Some games (both single-player and massive multiplayer) already include complex economic systems as part of their design mechanic. It would be easy to embed real-world, real-time financial data (such as commodity prices) within those games. Companies could track how players react to the data, then aggregate the reactions to predict real-world economic events accordingly. And, as massive multiplayer games with rich, dynamic economies become increasingly popular, opportunities to learn from player behavior will be enhanced accordingly.

Players Don’t Need to be Experts — Or Even Close

Games like these could enable companies to harness the wisdom of thousands (if not millions) of independent individuals, and harness it in a manner that reduces the risk of group-think (one of the few deadly sins in a communal, predictive system). And while it’s easy to worry that “average people” can’t contribute to the solution of complex problems, evidence suggests that Average Joe has more to offer than you might think; at least, in aggregate. That’s a central premise of the Wisdom of Crowds.

I encourage my readers not to fixate on any given example of this idea, which may or may not have specific flaws. Rather, consider the bigger picture and — if you’re so inclined and don’t completely disagree with my premise — propose examples that may work better than mine, in this context. Collective intelligence has enormous potential value — we just need the way and the will to unlock it. Video games offer an interesting potential mechanism via which to do just that.