The Proteus Paradox

by Nick Yee

  The player narratives in this chapter provide a very different perspective of how relationships work. Players developed relationships after going through adventures together and sharing in both disappointments and accomplishments. Relationships are created when two people are willing to work through life’s ups and downs together, whether online or offline and regardless of what their checklists look like. Lasting relationships are forged, not found. And the truth is that relationships are a lot of work. This is why online dating websites are so seductive. They are selling the myth that compatible relationships can be deduced with a magical database and require no real work on your part. For both good and bad, online games turn out to be a lot of work. And lasting relationships happen in online games because the gameplay often forces people to build trust and work together. These games jumpstart that forging process.11

  CHAPTER 8 TOOLS OF PERSUASION AND CONTROL

  In the last five chapters, I’ve described how offline categories—such as our ethnicity or gender—find a way to contaminate the assumed utopias of virtual worlds. But the Proteus Paradox isn’t just about how our offline politics don’t change when we enter virtual worlds, it’s also about how things beyond our control end up altering how we think and behave. In this and the remaining chapters, I’ll detail how virtual worlds provide unparalleled tools for controlling us.

  We’re all liars. We’re trained to lie to each other in very specific ways so that we immediately forgive each other. And the aisles of supermarkets and drugstores supply us with a vast arsenal of these tools of deception. There are toothpastes and peroxide strips to make your teeth whiter. There are hair-coloring products to cover your grays, and lotions and powders to smooth your complexion. And there are pushup bras and compression shirts to enhance your appearance. As a society, we not only tolerate these modest and temporary self-transformations, we encourage them. No one will get mad at you because your teeth are a little whiter. By conforming to these ideals of appearance, we create a favorable impression on people we meet and interact with. We call it “dressing to impress” or “putting your best face forward.” But what does it mean to put your best face forward in a virtual world in which you can swap faces at the click of a mouse button? Virtual worlds like Second Life feature an extreme form of customization; as the company’s product factsheet states, “Using over 150 unique sliders, change everything from your foot size to your eye color to the cut of your shirt.” What are the consequences of being in a world in which facial reconstructive surgery is neither expensive nor painful but cheap and instantaneous?1

  Stealing Faces

  When you play a video game on a screen with a resolution of 1900 × 1200 pixels, your eyes receive over 2 million pixels every second. Amazingly, not only do our brains process that information continuously, but gamers can react and respond with lightning reflexes. To accomplish this feat, our brains highlight and prioritize the processing of patterns or points of interest. Faces are a good example of these privileged patterns. Even babies, ten minutes after being born, pay more attention to images that resemble a face than scrambled versions of those same images. Their brains quickly pick out any visual patterns that resemble an oval with two eyes, a nose, and a mouth placed appropriately.2

  Our brains employ many other heuristics to sort through the information deluge of everyday life almost instantaneously. After we pick out faces in this visual data stream, we next need to decide how to react to those faces, the bodies they are attached to, and the gestures their hands are producing. When we meet a stranger at a social gathering, we don’t pause the conversation for a minute to sort through this information consciously and then weigh the pros and cons of whether this is someone we want to meet. Instead, a gut feeling drives our interactions with this stranger. One cue that our brains use to make these snap judgments is how similar the other person is to us. We perceive people who look like us or believe in the same things we do as more attractive and more persuasive than people with whom we share little.3

  Our brains don’t have much time to make these decisions, so it turns out that even entirely arbitrary points of similarity can make us like another person more. In a study conducted by psychologist Jerry Burger and his colleagues, undergraduate students met another student taking part in the study—actually a research assistant posing as a participant (a “confederate” in experimental design jargon). During the study, half of the students discovered that they “coincidentally” shared the same birthday with the confederate. The other half, the control group, did not share the same birthday. After the study ostensibly ended, the confederate asked each student if, as part of a class assignment, he or she would mind reading an eight-page essay and providing a one-page critique in just twenty-four hours. The students who thought that they shared the same birthday with the confederate were almost twice as likely to agree to this burdensome request than were those in the control group. Using the same study framework, Burger and his colleagues found that the same effect could be elicited if the students were led to believe that they shared the same first name or fingerprint patterns with the confederate.4

  If faces are given priority treatment by our brains, and similarity leads to attraction and cooperation, could we leverage the extreme customization of virtual worlds, not to create new faces, but to create strangers that look like you? Instead of stealing your birthday or your first name, these virtual strangers could literally steal your face. Of course, this transformation is impossible in the physical world. You could dye your hair or wear colored contacts, but there is no easy way to change your facial bone structure to match another person’s face. In the digital world, however, this is an easy problem to solve. Commercially available software allows you to blend two digital photographs by indicating points of reference—the corners of the eye, the hairline, and so on. This makes it possible to create a virtual stranger whose face shares a precise 20 percent or 40 percent resemblance to your face.

  On the eve of the 2004 presidential election between George W. Bush and John Kerry, my graduate adviser, Jeremy Bailenson, was chatting with another professor in the communication department with a background in political science, Shanto Iyengar, about face morphing technology. Because people are more attracted to and more likely to help someone who looks like them, they wondered if they could influence how people voted in an election by blending people’s faces into well-known political candidates. In short, if a political candidate looked a little more like you, would you be more likely to vote for him or her? Another graduate student, Nathan Collins, and I were brought onboard the project. Even though past studies suggested that similarity cues increased attraction and persuasion, it wasn’t at all clear that it would matter in a high-stakes situation like a presidential election in which people had strong feelings about the candidates. And given how often people had seen posters and television ads of both candidates, it also wasn’t clear whether we could tamper with their faces at all without people noticing immediately.

  We decided to hedge our bets. We would use a conservative 20 percent face morph for half the study sample and a higher-risk 40 percent morph for the other half. The 160 study participants were a nationally representative sample of voting-age US citizens recruited by Knowledge Networks, a company that offers free web TV in exchange for regular participation in online surveys. The participants were first asked to provide a digital photograph of themselves to allow us to create the morphed candidate images. Then, one week before the presidential election and ostensibly as an entirely separate study, the same participants were shown a photograph of Bush and a photograph of Kerry and asked about their opinions of the candidates and for whom they were likely to vote. Each participant saw one candidate morphed with their face and the other candidate morphed with someone else (to control for any extraneous effects of digital morphing). We found that face morphing made no difference to strong partisans, but weak partisans and independents were swayed by it. Overall, the effect was significant: among the group of particip
ants morphed with Kerry, Kerry received enough votes to win the presidential election.5

  This study was one of four studies we conducted using face morphing. Altogether, more than six hundred participants were shown digitally morphed photographs blending their faces into political candidates, ranging from 20 percent to 40 percent morphs. In each study, every participant was asked to guess the goal of the experiment. Roughly 3 percent of participants believed that someone had manipulated the candidate’s image. Not one participant across the four studies suggested the possibility that his or her own photograph had been blended with the political candidate’s. In virtual worlds, tools of persuasion can be powerful yet undetectable.6

  Breaking Reality

  This strange new world of face stealing was where I found myself in 2004, a year into my graduate program at Stanford. For four years, I worked with Jeremy Bailenson in the Virtual Human Interaction Lab on dozens of studies exploring one question: What if you could break reality? This question is rooted in a peculiar feature of virtual worlds. Virtual worlds do such a good job of creating the illusion of shared reality that you have to take a step back to remember that a computer system mediates all the interactions between users. Everything that two people say to each other has to first pass through this digital intermediary. In a world in which two people never have direct contact with each other, the messenger is king. And if you are in control of the virtual world, the messenger is working for you, and thus you become the king.

  The implications are far-reaching. In the most simplistic scenario, imagine that the system performs basic sentence parsing and makes sure that all requests contain either the word please or a smiley face. If it doesn’t, the system simply injects one or the other. Player A is unaware of the injection, and Player B responds more positively than he or she might have, starting off a positive feedback loop. Or these superpowers could be the king’s privilege. The king could have a dozen automated helpers injecting mundane pleasantries as needed. The king is always smiling, always saying, “Thank you,” and always remembers the names of your spouse, children, and favorite sports team.

  But the king’s powers go well beyond that. Consider eye gaze. We know that students feel more engaged and learn better if the teacher pays them more attention. In a virtual classroom, each student sees the shared reality of the virtual world only via his or her own individual display device. This means that these many slices of shared reality do not need to line up with each other. Thus, the teacher in a virtual classroom can maintain eye contact with every single student at the same time. Each student would feel that he or she has the undivided attention of the teacher. In fact, in a virtual classroom, you can warp space itself. In a physical classroom, only one student can sit in the front center of the classroom. In a virtual classroom, there is no reason why every student cannot sit in the front center of the classroom, each perceiving the other students dynamically repositioned around him or her.

  We also know that an unruly student can distract other students in a classroom. In a virtual classroom, the king can automatically censor the unruly behavior, splicing in a loop of recorded good behavior from a moment ago and projecting that instead to the other students. By placing students in a roomful of perfectly attentive students, we increase the likelihood that each student conforms to that model behavior. These superpowers could be combined to create the perfect virtual classroom. Indeed, my colleagues and I at Stanford have created these very virtual classrooms, placed students in them, and found that they improved learning.7

  This broken reality also reveals the true power of face stealing. In a virtual world, a presenter’s avatar could be individually blended with each audience member. Audience members each see their own version of reality in which the presenter looks subtly like them. In virtual worlds, political candidates can literally have a thousand faces. And of course, face morphing is only one of many possibly transformations. A benevolent king could create the perfect classroom, but a more devious king might create persuasion chambers tailored to each audience member. The devious king would steal part of your face and maintain eye contact with you while you sit front and center. And of course, the king’s teeth will also be that impossible shade of brilliant white.

  The Proteus Effect

  In 1966, psychologist Stuart Valins conducted a lab experiment in which male undergraduate students were asked to look at models from a Playboy magazine while they were hooked up to a machine that amplified their heartbeat and made it audible. The students were told that the experiment was a study of physiological reactions to visual stimuli. Of course, because this was a psychological experiment, the machine wasn’t actually amplifying their heartbeat. It wasn’t recording anything at all. Instead, the heartbeat noises had been prerecorded and were being played back by the machine. When the students were looking at some of the models, they would hear their “heartbeat” increase noticeably. At the end of the study, the students were asked to rate the attractiveness of each model. The students rated models randomly paired with an increased heartbeat as being more attractive.

  But why should a bogus heartbeat influence how students rated the models? Wouldn’t the students have formed an impression of each model based solely on the photograph? Deciphering our own emotions and attitudes is not straightforward; we do not keep an up-to-date reference list of our attitudes toward every person, situation, or social issue we might encounter. In many cases, our own thoughts are a black box, even to ourselves. We do this self-deciphering without conscious thought because this is how we understand other people. We don’t have direct access to other people’s inner thoughts and must infer their attitudes based on how they behave. Whenever Sam comes home, the first thing he does is sit down in front of the TV and play video games. He probably likes video games. Whenever the art lesson starts, Rachel slumps in her chair and frowns. She probably doesn’t like art lessons. In the same way that we use other people’s behavior to infer their attitudes, we do the same with our own black box. We unconsciously and automatically observe our own behaviors to make sense of how we feel about something. Thus, in Valins’s study, the students notice their rapid heartbeat, and since the only stimulus in the room is the Playboy model, it must be the cause of their arousal. And if they are aroused, the model must be very attractive. This self-perception theory reverses our intuitive understanding of how our brains work. Our behavior isn’t directed by our attitudes. It’s the other way around. Our behaviors direct our attitudes.8

  Of course, real life is never as sterile as a psychology lab. The cause of our excitement or sadness at any given moment can be ambiguous and difficult to pinpoint, especially in the typical data deluge our brains receive. This leads to many interesting accidents. The Capilano Suspension Bridge is a 5-foot-wide, 450-foot-long pedestrian bridge that stretches over the Capilano River in North Vancouver, British Columbia, with a 230-foot drop to the rocks and rapids below. The low handrails and the bridge’s tendency to tilt and wobble as one crosses it create the constant impression that one is about to fall. In 1974, psychologists Donald Dutton and Arthur Aron conducted a study on the bridge specifically because of its “arousal-inducing features.” A solid wood bridge conveniently located upriver served as the control condition. On each of the two bridges, a female research assistant approached male tourists as they approached the midpoint of the bridge and asked them to fill out a questionnaire for her psychology class. When the tourists finished the questionnaire, the research assistant gave them her phone number and told them to call her later if they wanted to know more about the study. Tourists who met the female assistant on the suspension bridge were more likely to call her afterward than were the tourists who met her on the solid bridge. In Valins’s bogus heartbeat study, the students attributed their arousal to the photograph. In the bridge study, the arousal caused by crossing a shaky bridge was incorrectly attributed to presence of the female assistant. This caused the men to be more attracted to the female assistant and thus more likely to call her
later.9

  We decipher our own attitudes based not only on what we do but also on what we wear. In the mid-1980s, psychologists Mark Frank and Thomas Gilovich tabulated past records from the National Football League and the National Hockey League. They discovered that teams wearing black uniforms received more penalties than teams wearing uniforms of other colors. Because teams change uniform color depending on where they play, Frank and Gilovich were able to show that when the same team wore black uniforms, it received more penalties. They also found that people perceive players in black uniforms as being more aggressive on the field. This led them to wonder if wearing a black uniform made players more aggressive via self-perception. To exclude the possibility of referee bias, they conducted a lab experiment. They brought students into the lab in groups of three, randomly assigned each group either black or white uniforms, and led the students to believe that they would be competing against another team of students preparing in another room. Each group then saw a list of twelve games from which to choose the five they wished to compete against the other team in. Frank and Gilovich found that study participants given black uniforms selected more aggressive games than participants given white uniforms. In their words, “Just as observers see those in black uniforms as tough, mean, and aggressive, so too does the person wearing that uniform.”10

  By 2005, I had worked with Bailenson for two years, exploring how transformations in virtual worlds could grant social superpowers—how a digital transformation could influence people’s interactions. I began to wonder if there was anything to the flipside of that question. Your avatar in a virtual world or online game is like a super uniform. It is infinitely more fluid and encompassing than a black outfit. Your age, your gender, and your body proportions are all things that are easily modified in a virtual world. And if a simple black uniform can make someone more aggressive in a laboratory, what happens to you when you are put into a digital avatar?