Game on: Video games help test human behavior
Studies analyze in-game economies, how long players play and how players deal with in-game infections.
Mon, Dec 20, 2010 at 09:59 AM
TEAMWORK: Researchers have collected data about the overall behavioral patterns of thousands of gamers who play together in the online game Everquest II. (Image: Sony Online Entertainment)
As a communications researcher who focuses on virtual worlds, Dmitri Williams has a simple rule for how his team studies games: If you haven't played it, you can't study it. As a gamer, it sounds like a no-brainer.
But when I voiced my admiration for the rule and its sensible approach, Williams pointed out that I'm the "first journalist ever" to not ask him whether he plays video games. His usual stock response to such questions: "Would you ask film professors if they watch films?"
Williams is based at the University of Southern California, though he heads the Virtual World Exploratorium (VWE) project that spans multiple universities. His group of economists, psychologists and anthropologists see games as nothing less than virtual labs where they can collect data about human behavior that would be impossible to get in the real world. [Understanding the Most Destructive Human Behaviors]
"The worlds are great petri dishes for studying human behavior," Williams said. "Most of us are not interested in a particular virtual world. We don't care about the dragon, we care about human behavior."
The group's ability to mine useful data is a big reason why it has funding from the National Science Foundation, the U.S. Army Research Institute and the Intelligence Advanced Research Projects Activity – the latter being the high-risk, high-payoff research arm for the U.S. intelligence community.
Every move you make
As inspiration for his work, Williams pointed to the 1966 book "Unobtrusive Measures" by Eugene Webb, a pioneer in organizational behavior research. The book tells of a museum that wanted to measure visitor interest in certain exhibits without actually bothering visitors, and so it used the ingenious solution of examining the wear and tear on the floor tiles in front of displays.
Similarly, players can do what they like in games without being bothered by monitoring devices or cameras, because their behaviors are recorded unobtrusively as in-game data. The big difference is that games can offer a much finer level of detailed behavioral data than any scuff marks found on a floor tile.
"The advantage of behavioral data is that rather than asking someone what they did, you know what they did," Williams told LiveScience. "It's the difference between asking what a person saw at the museum and watching what they looked at on the [camera] video feed."
Using game data also eliminates other real-world challenges that researchers face when trying to study human behavior. There is no need for survey data tainted by lying or poor memory, because researchers can simply consult game logs to find out how much a player really paid for a pricey dragon mount, or see just how long players spent playing online. (Hint: its more than players think.)
But researchers can only get their hands on such data if they approach game companies. Williams and his colleagues at the VWE project struck such a deal with Sony Online Entertainment, a game publisher that operates many massively multiplayer online (MMO) games such as EverQuest II.
To address possible "Big Brother" concerns, both the researchers and Sony Online Entertainment agreed that the player data would be anonymous. That means the researchers don't ever see any personal identifying information connected to the game data. Since then, the VWE has also struck similar deals with a Chinese MMO publisher and a U.S. free-to-play online game publisher.
The deal has already paid dividends for the researchers, who hail from diverse academic backgrounds. One economist wrote a paper about macroeconomic behavior in a virtual world — not unlike how other experts have looked to online worlds such as "Second Life" for lessons about unregulated banks and economic patterns.
"When we look at a virtual world, we have every single transaction, period, with no error – there's no sample because we have everything," Williams said. "We have every [virtual] copper piece that was ever sold for a wolf pelt." (Williams is referring to the option players have in many games of buying/selling/trading with either a non-player character or other players.)
Such economic research not only benefits from a virtual world's ability to track every purchase of every wolf pelt, but can also examine unusual economic circumstances that play out in a virtual world with no real-world costs. Policy makers don't have that luxury when they try to tackle real-world economic problems such as recessions.
"In a virtual world, you can test things with much less risk and at much lower cost than in the real world, where policy makers run these kinds of experiments all the time," Williams said.
Another colleague with an interest in transportation patterns studied why players chose certain routes based on travel times. That study relied upon the basic psychology of decision-making that underpins what humans choose to do, even in virtual worlds where players trade for fantasy items and ride mythical griffins or dragons from point A to B. [Related: Grand Theft Auto May Improve Decision-Making Skills]
Still, getting the research right means figuring out what in-game behaviors are a good match for behaviors in the real world. Researchers risk doing bad science, and ending up with inaccurate conclusions, if human behavior in a game "maps" poorly to real-world behavior.
"Mapping is all about when can I trust this parallelism and when can I not," Williams explained. "When does it go from goofy research with a bunch of professors studying dragons to a study with practical applications in the real world?"
He gave the example of a paper published in the journal The Lancet Infectious Diseases in 2007. That paper explored how researchers might study human behavioral reactions to epidemics in virtual worlds, and cited a much-publicized instance where a disease-like condition began spreading and killing players' virtual avatars in the popular "World of Warcraft."
The problem with the Warcraft example was that players likely behaved radically different from how they would have in a real-life situation. Infected players gleefully spread the disease to major cities in the game with predictably messy results, because virtual death in "World of Warcraft" is only an inconvenience.
"If players had a psychological cost associated with [in-game] death, like losing a week of playtime, maybe games would start to approximate real-world systems," Williams said.
Taking games seriously
Having a rigorous approach to studying games has helped the researchers gain credibility for their work, given that relatively few researchers look at games beyond their impacts on such things as aggressiveness and violence.
One of the first researchers to take games seriously was Nick Yee, an expert in the psychology of virtual environments at Xerox's Palo Alto Research Center in California. Back in his undergrad days in 2002, he began surveying gamers who played early MMOs such as EverQuest and publishing his results online.
"Nick was really one of the very first guys to look into this space," Williams said. "Here was an undergrad kid doing a lot of stuff that was state-of-the-art [in this field] for a long time."
But whereas much of that early pioneering work was focused on scraping publicly available data from games, Williams and other researchers have gone deeper into the rabbit hole by getting game data through the back door, so to speak. Their challenge now is to keep demonstrating the potential of virtual worlds as virtual labs, and open up new opportunities for working with the game industry.
"It's easy to blow it off, and it's also easy to get it wrong," Williams said. "We take this very seriously, we're funded by serious agencies and we publish in serious journals."
This article was reprinted with permission from LiveScience.
Related on LiveScience: