What a Video Game Can Teach Us About Getting Through a Pandemic
Anna Weltman on Making Real-World Models From World of Warcraft Epidemics
Nowhere is it more important to have a good real-life model than during an unpredicted, quickly evolving emergency, such as an epidemic. Epidemiologists rely on mathematical models of human behavior during epidemics to make time-sensitive decisions, such as where to send doctors and valuable supplies. But how will people act during epidemics? Will they behave as John Forbes Nash might predict, caring for themselves over others? Or will trust and relationships come into play? Will they see each other as competitors for scarce resources or as colleagues with mutual interests working toward the common good?
Mathematicians’ models for behavior during epidemics were put to the test in one of the largest epidemics ever: the Corrupted Blood Epidemic of 2005. Around four million were infected. It threw a complex society into chaos as officials scrambled to contain it. Eventually they succeeded, but not before many died and game theorists’ ideas about human behavior were irreversibly challenged.
Never heard of the Corrupted Blood Epidemic? That’s probably because you don’t play World of Warcraft, a computer game involving a host of imaginary creatures. Players from all over the world play against each other as these creatures. Corrupted Blood wasn’t a human epidemic. It was an epidemic accidentally released into the cyberworld by a software update gone awry. The epidemic was pure happenstance, at least as far as the humans involved with the game were concerned. As such, it provided the safest possible setting to test mathematicians’ and epidemiologists’ understandings of the ways people behave in a crisis.
On September 13, 2005, approximately four million World of Warcraft players were exposed to the full-blown epidemic. Corrupted Blood infected characters of all species. Orcs, dwarves, night elves, and pandaren, pandas with human hair, all succumbed. The players, glued to computer screens around the world, did not understand what was happening. Nothing like this had ever occurred in World of Warcraft. Some characters recovered, some died, and many others spread the disease to distant corners of the World of Warcraft universe. Even the game’s developers hadn’t anticipated the reach of the disease. Stopping the epidemic would require more than a few keystrokes.
Those unfamiliar with World of Warcraft might think that the Corrupted Blood Epidemic could be stopped if everyone just took a break from the game while the game’s developers found a fix. But most fans are die-hard enthusiasts. They couldn’t take that break. They had to heal their characters before it was too late. For them, a World of Warcraft epidemic was nearly as serious as an epidemic in the non-virtual world.
What would you do if an epidemic like Corrupted Blood were ravaging the real world? Lock your door and bar the windows? Horde supplies, leave friends and family to sink or swim, and protect your immediate household from the infected masses? Or risk your life trying to help those affected? What if one of the people needing help was a close relative across town?
These are precisely the questions that mathematicians Nina Fefferman and Eric Lofgren sought to answer by studying players’ reactions to the Corrupted Blood Epidemic. They published their findings in a respected medical journal, The Lancet. Medical journals don’t typically address virtual medical problems contracted through software updates. They certainly do not deal with illnesses endemic to virtual orcs and elves. But Fefferman and Lofgren’s findings were relevant to the journal’s main concern, human health. What bridged the two seemingly distant worlds of medicine and online games was math.
To mathematicians, how people behave during an epidemic is a game. An epidemic is on its face a competitive situation because the two groups of people involved—the sick and the healthy—can be seen as having opposite needs. They have the same basic goals, for the epidemic to end and to stay alive until it does. But their ways of achieving their goals could conflict with each other.
During an epidemic, it’s in the best interest of healthy people to stay far away from the sick. If healthy people stay away from the sick and don’t catch the disease, the epidemic is more likely to dissipate than if the healthy people continue to interact with the sick people. But it’s in the best interest of sick people to find a healthy person to take care of them. If sick people don’t get care, it is more likely they will die. Therefore “winning” for one type of person could be “losing” for the other.
Many mathematical models that predict what people will do during epidemics presume that people will act selfishly. The Nash equilibrium’s solution to the Prisoners’ Dilemma did the same. If the healthy and sick are competing to avoid dying from the disease, wouldn’t they each take the action that is best for themselves, regardless of what happens to those around them? The assumption is that people won’t care about the people they’re competing against—each person will act for themselves.
But in some situations that have competitive components, such as trying to stay alive during an epidemic, it turns out that how people behave is much more complicated. The problem is that classic game theory models don’t explain why healthy people sometimes risk their lives to care for the sick. It also doesn’t explain why some sick people separate themselves from others in heroic efforts to reduce the spread of disease. As it turns out, real people don’t always try to win. In real-life epidemics, as in the Prisoners’ Dilemma, relationships matter when people decide what to do.
And it’s important that mathematicians and epidemiologists know in advance what people will do. That’s because epidemics are spread by human contact. Public health officials need to be proactive to stop epidemics, so they need to know who will reach out to whom. If people behave as Nash predicts, with the healthy isolating themselves and the sick seeking care, the disease will take one trajectory. But if people try to help or protect those they care about, possibly including strangers, the disease may spread in a completely different way.
So, how do people act during epidemics? What factors affect whether people will act altruistically or selfishly? Perhaps it’s how awful the disease is. Maybe it’s how close people’s relationships are in the community in which the disease is spreading, or how medical expertise is distributed. These are important questions for mathematicians to answer if they’re going to build accurate computer models about epidemics. But it’s hard to study real people in these circumstances. So, Fefferman turned to World of Warcraft and its Corrupted Blood Epidemic for answers.
Fefferman, a professor in the biology and math departments at the University of Tennessee, uses math to model how people behave during epidemics. She typically collects data while epidemics are happening, examining the relationships between human behavior and the spread of disease using computer models. Sometimes Fefferman also uses models that other people developed to test claims about what might happen during a real epidemic. But her most groundbreaking research involves inventing models herself.
World of Warcraft was a good site for Fefferman’s behavioral research because the characters in the game could get sick—and die—without hurting any actual humans. But the characters’ behavior was completely controlled by players who cared about their characters’ health.
World of Warcraft was, of course, just a game. What happened in the game had no direct bearing on real life. Fortunately for the players, the company that makes the game was eventually able to stop the disease and reset much of the damage. But for many of the players, the epidemic was a real-life experience for them.
World of Warcraft players can be extremely serious about the game. They spend hours playing each week, taking their characters through elaborate quests that lead to personal glory and advance the plot of the game. In doing so, players develop relationships with each other. A sense of community develops in the game. During the Corrupted Blood Epidemic, the altruism that grew out of community attachment seems to have greatly changed the course of the disease.
Sick characters, under the control of their human players, traveled around making others sick—sometimes because they didn’t know about the risks, and sometimes because they were looking for help. Fefferman observed characters with special healing powers flocking to areas with lots of sick characters, trying to help. Some of the characters in World of Warcraft were less susceptible to Corrupted Blood than others, much as some humans are less likely to suffer from epidemics than others. But, like humans, characters who were less susceptible to dying from the disease but were carriers of it kept accidentally spreading the disease to weaker characters—especially when they rushed in to save those in need.
As a result, altruism didn’t always help. Actions that may have seemed logical to a player trying to help his friend had implications beyond what anyone anticipated. In turn, altruism affected the spread of the disease in ways that led Fefferman to conclude that traditional game theory models weren’t accurate. Sometimes altruism made things better, sometimes worse. In any case, game theorists needed to factor altruism into their models. Fefferman made new models that captured what she had learned from the Corrupted Blood Epidemic. Those new models are now in use around the world, helping us prepare for real epidemics when they strike.
But what Fefferman’s work shows more than anything else is that in real imperfect-information games, ones that are far more complex than the Prisoners’ Dilemma, mathematicians still have a lot to learn. When human behavior takes mathematicians by surprise, maybe the humans aren’t the ones acting illogically. Maybe it’s the mathematicians who don’t understand.
From Supermath by Anna Weltman. Used with the permission of Johns Hopkins University Press. Copyright © 2020 by Anna Weltman.