Recently, my friend Penny sent me a link to a brief news story about racial bias in There.com, reporting on a study conducted by Paul W. Eastwick and Wendi L. Gardner, psychologists at Northwestern University. The experiments are outlined a bit more clearly in another piece I subsequently dug up, but in both reports they're easily recognized as testing classic social psychological findings—or at least, easily recognized by those of us so accustomed to social psychology research that our first reaction to seeing a not-obviously injured person lying next to a bicycle on a campus sidewalk is to wonder, Is someone doing an experiment on helping behavior? Basically, the researchers entered public spaces in the socially-oriented virtual world of There, approached randomly selected avatars, and performed one of three possible experimental treatments:

1. Made an unreasonably intense request: "Would you teleport to this particular zone so I can take a screenshot of our avatars there?"
2. Made a very unreasonable request ("Would you teleport with me to 50 different locations to take screenshots?") followed by the only somewhat unreasonable request from Condition #1. (Also known as a "Door in the Face" situation.)
3. Made a very moderate request ("Can I take a screenshot of our avatars here?"), followed by the unreasonable request from Condition #1. (Also known as a "Foot in the Door" situation.)

As expected, in Condition #2, the Door in the Face, participants were more likely to acquiesce to the only somewhat unreasonable request than in condition #1—refusing a huge request makes people conscious of the social desire for compromise when the smaller one is made. In Condition #3, the Foot in the Door, participants were more likely to acquiesce to the unreasonable request than in condition #1, because they had entered into a "helping frame of mind" by granting the first request.

The experiment is made more interesting by the fact that there were actually six possible treatments, however, as each treatment could be conducted by an avatar with the lightest of the three skin tones available to free users in There, or by an avatar with the darkest of those skin tones. Avatars with dark skin were significantly less likely to obtain acquiescence to the only somewhat unreasonable request in condition #2. In condition #3, there was no difference; as the second article linked above points out, people's desire to compromise is influenced by their perception of the person doing the requesting—if that person is not someone they look at as Someone To Be Impressed, then they are less likely to feel the social pressure to compromise. In contrast, perceiving oneself as being Someone Who Is Helpful is not related to the person being helped. The results of these experiments in There confirm both previous social psychological theories about inducing compliance with requests as well as theories about how racial bias influences interactions between strangers.

As a researcher who firmly believes that there are more similarities than differences between social interaction online and social interaction face-to-face, and whose own research in fact hinges on the assumption that classical social theory will be born out in virtual interaction, it's nice to see some confirmation. As an instructor for a course on race and ethnicity in American social life, it is depressing but not particularly surprising to read three pages of reader comments along these basic lines:

The only thing bias in this article is the article itself. Seems to me the person conducting this research has no clue about on line communities especially There. The only thing you can equate to the real world is that no one is going to let a complete stranger take pictures of them and teleport around with them ¹. Doesn't matter what sex or color the avatar is. There are island guides for people who need help when they come into There. To approach people asking them to take time away from their game play to let you snap pictures of them and teleport around with them border lines on creepy. With that being said how valid is your research? If your avatar was blue, purple, white or green you would of gotten the same responses. Next time get to know the culture of a virtual world before you print lies like this. Who ever published you also needs to get a clue.

While the news stories are sadly a bit thin on the study's exact sample selection methods, and my university library cannot provide me with timely access to the full journal article, it's hard to believe that they would have gotten published without basic randomization, which is central to all social psychological experiments. Administering experimental treatments at random, light-skinned avatars received a 20% increase in acquiescence in Condition #2, while dark-skinned avatars increased acquiescence by only 8%. We could certainly have more or less confidence in the results if we knew the exact number of samples, but the advantage of performing experiments in a virtual environment is that one can probably get a bigger number rather than a smaller one. Bottom line: there was a large and significant difference between "Door in the Face" compliance with light-skinned avatars and with dark-skinned ones. There's no social scientific reason to think these results are invalid—and I say that as a researcher who frequently complains herself about other researchers' lack of native understanding of technology use of various kind. In this case, lack of conformity to There.com norms does not invalidate the findings.

The comment criticisms of the study seem to me to be indicative of a general tendency among white people—and let's not imagine that the vast majority of There.com users are not white people—to refuse to see racism in the world no matter what evidence is presented. This may be conflated with the old techno-utopian dream of the internet as a magical place where bodies don't matter, or even the old favorite, that virtual bodies don't matter and of course no one ever acts as if they do. Another commenter argues, "The test is flawed because you do not know what colour skin person is behind the avatar," as if no one would ever treat a black avatar like a black person, whatever kind of treatment that might indicate. Of course people do generally treat female avatars as if they were female people, despite widespread knowledge of gender-swapping; some gender-swapping users hoping for various kinds of assistance are banking on this. Gender is generally more variable than race within virtual worlds, but the two categories are alike in their broad use as status indicators. My own experience in City of Heroes has been that black bodies are subject to scrutiny in a way that white bodies are not: an afro hairstyle is cause for comment but no one ever says anything about a pink-haired superheroine. There.com, in fact, is the only virtual environment in which I have ever been explicitly queried about my own race. It did strike me that There users were vastly more concerned about "authenticity," i.e. as-perfect-as-possible correspondence between avatar and user, than, for instance, Second Life users—gender-swapping was uncommon due to the prevalence of voice chat, for instance. My very pale, purple-haired avatar was never asked if her hair was accurate, but I remember a specific occasion when a white male avatar, who seemed to be chatting me up, asked me if I was "really white" and reacted defensively when I indicated that I found the question irrelevant to our interaction.

So as I have argued before, race matters in the virtual world, and in fact the general invisibility of people of color in most virtual worlds may heighten differential treatment by perceived differences in race. Statuses like gender and race are generally used to make judgments about others the most when they maximally discriminate between individuals—that is, being black doesn't hurt you when everyone else is, too, or even if there's a relatively even mix, but if you're one of only a few black people in the crowd, you are likely to be singled out by that status—which, of course, contributes to the sense of the world as one where black people do not belong. Furthermore, there is a general sense in the everyday face-to-face world that race is a kind of commodity, that people choose to mark themselves racially and could be "normal" if they would simply give up their exotic fashion statements: dreadlocks, saris, hair that has not been straightened to look appropriately white . . . In a virtual world, where the "cost" of wearing only markers of whiteness is seen as especially low, of course people who mark themselves otherwise should expect to be singled out in various ways—they're asking for it.

One question I do have about this work, beyond basic issues of methods and sample size too boring to report in the popular science press, is why academics continue to focus inquiries like this one on the social world platforms like Second Life and There.com. I would actually argue that There is a better venue than Second Life, by virtue of its lower technical specs and somewhat less homogeneous population—while there is certainly variation in the demographic characteristics of Second Life users, most of them are still, in some way, members of a cultural and technological elite who possess the resources necessary to run Second Life without kicking their computers to death in frustration. The media love Second Life—hell, it was featured prominently on CSI:NY—but it makes it hard to determine how many accounts get started and barely used by people who thought it looked cool but aren't really ready for the high technical specs or steep learning curve. Wikipedia reports 15 million registered accounts on Second Life as of September 2008, but only a very small percentage of those are paid; in contrast, World of Warcraft has 10 million subscribers worldwide, all of whom are paying a monthly fee by definition. Some of them may still not play much, but it's a safe bet that they're playing more than the majority of those 15 million registered Second Life users.

Furthermore, while I hate to agree with the commenter quoted above who fails to grasp the basic advantage of randomized experiments, there is a basic issue of wanting to make an experiment "look" as much like Real Life as is manageable. It would seem reasonable that a goal-oriented massively multiplayer online game is a better environment to that purpose than a social platform with an underlying value, more or less commonly held/articulated, of "community." ²

It's basically true that when someone asks to take a screenshot of our avatars together, I'm not sure why they're asking. In meatspace, when people ask to take my photo, I know that it's because I have pink hair (and because I am vain—note the self-direction of my motivation—I am likely to acquiesce); in There, on the other hand, it's a rare avatar that really stands out enough in the environment to make the request sensible. Conversely, if someone in City of Heroes asks me for in-game currency (a request with which it was impossible to comply when I was using There, as the developers had purposely disabled financial transfers), I may be annoyed at their virtual panhandling, but I can easily imagine why they want some money. Similarly, I can see why a low-level character might ask me for help with a quest or what-have-you. If anything, though, I would expect results like those discussed here to be more striking in the game environment, because actions like assisting a lowbie on a quest are likely to slow my own progress in the game insofar as they reduce the number of experience points I'm earning for my time investment. Competition theories of racism emphasize group membership and the perception that one's group is competing with other groups for limited resources; games seem to provide a better framework for testing related behaviors. World of Warcraft has various non-human races, true, but experimenters are free to restrict themselves to humans and adjust skin color accordingly.

In fact, games, which are more structured than Second Life in terms of available bodies as well as offering a more goal-oriented environment, offer some interesting ways to test more general social psychological theories of group dynamics. Demonstrating that race is still used to judge others in virtual worlds is a very important finding, but what about the completely artificial races to which players can belong in World of Warcraft? Do Tauren feel in-group loyalty towards other Tauren? Is there in-group loyalty for particular classes? Is it more pronounced for some (paladins) than others (rogues)? Such experiments would, of course, raise the question of how much people really "identify" with avatars, but research suggests that it's really pretty easy to induce a sense of group membership in most people. Studies like the one conducted by Eastwick and Gardner are cool and important, and they demonstrate that yes, we can extend theory from face-to-face to virtual interaction. That said, what might we learn in the virtual world that is difficult or impossible to test in meatspace?

[1] While it is true that I would not allow a complete stranger to teleport me, I have in fact allowed a number of complete strangers to take my photograph. When you have hot pink hair, it is a not uncommon request. And let's be honest; I probably would let a stranger teleport me, because I have more than a passing interest in a) wizards and b) time-travelers/aliens.

[2] Second Life seems like it would produce especially small indications of bias since users seem to largely espouse particular kinds of socially liberal views—which is not to say that users would not still hold some racist attitudes, conscious or not, but that they would likely be more sensitive to them.