In the infamous Milgram’s Experiment of the 1960s, test subjects were placed in front of an intercom system and asked to administer increasingly painful electric shocks to a man in another room who failed to answer test questions properly.
The “victim” was, in fact, an actor working for Yale University psychologist Stanley Milgram. But screams coming through the intercom sounded real enough — and consequently, were incredibly traumatizing to the volunteer test subjects.
What if the victim was clearly not real, but virtual? This question is not a purely academic one, as some of tech’s very largest companies are aggressively introducing artificial intelligence-driven personal assistants and other avatars that will inform, entertain, and even befriend users.
The avatars we see in games and other entertainment may seem like harmless digital representations, but they have the real power to shape human perceptions and attitudes as we interact with them. Several studies suggest their impact on us can be profound — both for good and bad. And their pervasiveness in technology, especially in the COVID-19 era, promises an even deeper effect.
As a result, it’s crucial that technologists and business leaders understand how reality is changing and that they create and adhere to ethical guidelines and safety standards that will be required in a world we share with AI-driven avatars.
A fine line between real and imagined
To better understand the impact of avatars on the human mind, Microsoft Research’s Mar Gonzalez-Franco and the University of Barcelona’s Mel Slater recently recreated Milgram’s experiment in virtual reality and measured its effects on volunteers.
In their version, rather than being ordered to shock a human victim in another room, test subjects were told to electrocute a 3D avatar displayed on a headset whenever the subject appeared to answer a test question incorrectly.
Slater and Gonzalez-Franco discovered an amazing result: While knowing full well that the avatars were completely virtual, volunteers still were traumatized at the prospect of applying shocks to it, and some volunteers even attempted to help the avatar cheat on the test.
"An avatar that looks like you will probably be as practical as having an email address or a phone number.” — Lauren Bigelow, IMVU
Gonzalez-Franco believes the immersive power of VR makes it possible for us to experience empathy for a virtual being. Specifically, through VR’s ability to create a sense of presence and place.
“There is a strong presence illusion that enables the empathy mechanisms,” as she puts it, ascribing it to two factors: the “plausibility illusion that the things that are happening are realistic. And the place illusion, the illusion that you are in a new location, in this case the examination room.”
For Gonzalez-Franco’s iteration of Milgram’s experiment, several other factors enhance the illusion: The avatar’s movements were motion captured from a real actor’s performance, while the avatar’s dialog was lip-synced. Perhaps most importantly, the avatar was programmed to make eye contact with the user.
The experiment suggests that avatars appear real to human observers. Or, at the very least, that they can be made to appear sufficiently real in the moment to alter our reactions to them, the surrounding environment, and our perceptions of ourselves.
Over the past decade, academics experimenting on a wide array of platforms have produced multiple studies that bolster this belief. Many were conducted in the heart of Silicon Valley at Stanford University, but findings are not widely acknowledged in the tech world, even if the industry is already engulfed by avatars.
The use of avatars in online gaming, a single tech sector, is a case in point. Counting just three of the market leaders — Fortnite, Minecraft, and ROBLOX — about 500 million people interact with 3D avatars on a regular basis, often with more intensity of engagement than they do with television or other traditional media.
At a conservative estimate, there are dozens of avatar-oriented games that attract tens of millions of active users. Notably, the most popular of these avatar-driven game worlds are predominantly populated by people in their teens and early 20s, whose activity will shape their expectations of the Internet as they grow into adulthood.
Gaming is far from the only sector that will be influenced by avatars or AI-powered digital humans. Beyond their value in virtual reality, digital agents will likely take on a more central role in our Internet usage, especially for enterprise and therapeutic use cases in the post-pandemic age.
As this shift occurs, many new opportunities for developers and investors will emerge... as will the unintended hazards that could alter our social structures for the worse.
What follows is just a small sampling of a world remade by avatars, which will not only influence our entertainment but shape how humans feel about themselves and each other. Consequently, they will offer new opportunities to enrich our world for the better — or if we are not careful, for worse.
“The immersive power of VR makes it possible for us to experience empathy for a virtual being.” — Mar Gonzalez-Franco, Microsoft Research
Virtual inputs, real outcomes
As the latest wave of consumer models of VR headsets came to market, many included an interest clause in their Terms of Service agreements.
HTC Vive’s safety and regulatory section, for example, says: “It is important to remember that simulated objects, such as furniture, that may be encountered while using the product do not exist in the real world, and injuries may result when interacting with those simulated objects as if they were real, for example, by attempting to sit down on a virtual chair.”
Philosopher Rene Descartes wondered if the world he seemed to perceive was actually an illusion created by a malicious demon. Hundreds of years later, virtual reality products now carry what is effectively an “evil deceiver” clause in their terms of service.
VR’s effects may be even more powerful, as the technology has an ability to change the user in tangible, real-world ways.
Post-doc researcher Domna Bankou and her colleagues at the University of Barcelona, for instance, recently created a VR simulation where the test subjects would see themselves as Albert Einstein, standing authoritatively in a lab coat. They did this, Bankou says, with assumptions about what would then transpire.
“To be honest with you we weren’t expecting anything to happen,” she said. “But VR changes where you are, [and] it also changes who you are.”
In fact, after volunteers completed the Barcelona VR experiment and took a cognitive performance test, those with lower IQs saw their test scores substantially improve. The act of virtually embodying the smartest human in history apparently made them more intelligent in the real world — at least temporarily.
Bankou and her colleagues found similar results when white test subjects embodied a Black avatar, registering significantly less racial bias in a subsequent test.
The underlying reality of non-VR avatars
While Gonzalez-Franco and Bankou believe the effects of their experiments are mainly attributable to full body VR, it’s notable that we see a similar phenomenon in 2D virtual worlds.
In 2006, when virtual world graphics were far less realistic than they are now, Nick Yee, then a Stanford Ph.D. candidate, conducted a study that suggested people using a virtual world unconsciously follow humans’ unwritten rules of social distance and eye contact. In other words, the digital avatars maintained the same simulated level of space and duration of eye contact that humans would in a real-world situation.
This is important to note because, in the short term, people are more likely to experience virtual worlds on the flat screens of mobile, PC, and console devices. While a subset of hardcore gamers prefers highly photorealistic characters like those found in console and PC-based games such as Grand Theft Auto or Red Dead Redemption, the overwhelming preference among consumers doesn’t seem to be driven by immersive graphics.
IMVU, the company behind a mobile-focused virtual world of the same name, recently conducted a survey of 19,000 people between the ages of 13 and 30, asking about their preference for avatars in a social game space. Their choices seem to be driven less by graphics, and more by age range and the needs of a particular user.
“The most realistic avatar -- similar to the avatar style of IMVU -- was the top preference across all age groups we surveyed,” says Lauren Bigelow, Chief Product Officer at IMVU. “We also found that among younger respondents, the most cartoonish avatar was more likely to be their second choice.”
The difference in avatar preference varied widely by age range, even when the range was just a few years.
“13 year-olds were much more likely to prefer the cartoonish avatar as their second choice, but 17 year-olds were much less likely,” Bigelow says. “As teens mature they try to sort out who they are and establish their identity. The older they get into the teen years the more sensitive to being portrayed as ‘young” or ‘babyish’. This preference is based on a basic developmental reason but it’s not immutable.”
For instance, Bigelow notes that a sudden popularity in cartoonish avatars can break through this preference by age range.
One possible example for that is the massive surge in popularity of Fortnite among teens and people in their early 20s, despite -- or perhaps because -- avatars are decidedly cartoonish, with large expressive eyes and exaggerated facial features and bodies.
IMVU’s survey also found that users’ real-life social connections greatly shaped avatar preference. “In a majority of cases,” Bigelow says, “our users told us that they wanted their avatars to look like their real selves so their friends would recognize them in this virtual world.”
“VR changes where you are, [and] it also changes who you are.” — Domna Bankou, Post-doc researcher, University of Barcelona
Avatar applications in the age of COVID
While gaming and VR have done much to innovate the use of avatars in entertainment contexts, we will likely see them applied into serious, enterprise-facing use cases. This is especially true in the wake of the COVID-19 pandemic, with businesses and organizations actively searching for viable alternatives to web video conferencing that can convey a sense of physical preference and “being there” that the real world can no longer easily or safely offer.
A number of startups with high-profile clients and partnerships offer avatar-driven replacements, ranging from business conferences to music concerts. At the same time, innovations in robotics have opened up the possibility of real-world avatars -- that is, robots remotely powered by a person or even artificial intelligence.
Jacki Morie, a VR pioneer who advised XPRIZE's recent challenge to create a real-life robotics and VR-powered avatar, envisions a near future where we will use robotics-driven avatars to visit and interact with remote friends and relatives and "touch" them through Internet-enabled haptics devices.
That technology itself already exists. It is already possible, Morie notes, to use haptic-driven gloves that can command a robot in a distant location to reach out and grab something — and then “feel” the surface texture of the object that the robotic hand is holding.
Meanwhile, Lauren Bigelow at IMVU sees a future when avatars will become essential to our Internet identities. “As our experiences become more virtual,” she says. “it will be much more likely that we will need an avatar to represent ourselves, express ourselves, and interact spatially with other avatars. An avatar that looks like you will probably be as practical as having an email address or a phone number.”
Unlike a phone number, however, future avatars will be far more expressive of who a person is and what visual metaphor resonates with them most. “The power of a virtual space is that you can transcend the boundaries of real life,” Bigelow says. Depending on the situation, “you will want to express yourself more dramatically… and you might even want to have an alternate figure altogether to represent you like an animal, dragon, or flying superhero.”
As for the Milgram experiment with avatars conducted by Gonzalez-Franco, she sees many practical use cases that can be developed from its results. For instance, she foresees a simulation that could test the response of peacekeepers ordered to break-up a political demonstration that could encourage them to personally support and empathize with the demonstrators.
Gonzalez-Franco also sees great potential for VR to prepare users for consent in painful real-world situations. That could include a virtual reality scenario where someone undergoes cancer treatment or a clinical trial.
“The type of consent one provides before a surgical intervention or a clinical trial for cancer treatment is currently very much an abstract legal process, where you don’t really get to see what it means,” she says. “In that context VR could really bridge the gap: This is what can happen to you.”
In the end, this suggests a way of thinking about the power and potential of avatars: While they might only seem like a fanciful escape for entertainment, they can also help us become better versions of our selves — and better prepare us for the hard realities that will always exist behind the screens where our avatars live.