It’s no game, working to contain epidemics such as the flu outbreak that has been widespread or severe in 47 states this winter. Or is it?
In an effort to learn more about what motivates people to protect themselves from infection, a team of Wake Forest University researchers is using an online game that simulates the spread of an infectious disease among its players. The project – by economists Fred Chen, Amanda Griffith and Allin Cottrell, and computer scientist Yue-Ling Wong – is “ultimately a study of human behavior,” Chen said.
The team is excited about the study because it shows how people act in various decision-making situations, such as whether to get a flu shot, as opposed to relying on what they say about how they’re going to act. Wake Forest says it’s the first study in the economics of disease control using virtual diseases.
When studying how best to contain epidemics, scientists and policymakers often must make assumptions about how many people will or won’t take preventive measures to keep from getting sick. The virtual epidemic experiment shows how people really behave when faced with choices about whether to self-protect during a widespread infection in a community.
Researchers found in the baseline study that “players were rolling the dice to see if they could stay healthy without paying the costs of protection,” Chen said. “But even those players who were more inclined to take risks chose to self-protect the more often they got sick.”
Those results – which wouldn’t come as a surprise to many – serve a purpose: “I want to do a lot of experiments using this virtual disease framework,” Chen said. “This is the first set of experiments I ran, and I wanted to show that this is a valid protocol. If I had gotten something weird or totally unexpected, people might think that maybe this wasn’t a great framework for doing experiments. …
“The fact that we have very intuitive results is actually a good thing because it further validates our experimental framework. Although there are no experiments running currently, we are applying for funding to run more experiments.”
The risk-taking and the increased precautions after getting sick might not surprise students of human nature, but there were surprises from an economic standpoint, Chen said.
“The results we got are not fully consistent with standard economic models of decision-making,” he said. “In standard economic models, what we have is the following: If you are perfectly informed of all the rules of the game, your past experiences should not matter in terms of how you behave in the future. … Yet in the game, we can see that their experiences earlier in the game matter for how they behave in the future.”
How it was set up
The baseline study recruited players via Craigslist advertisements. “We do recruit online because we do offer compensation for participating in these experiments,” Chen said.
The game simulates an epidemic among the players for several weeks. At the beginning of each day of the game, healthy players have the option to choose, at a cost, a protective action that reduces the likelihood of getting infected.
Griffith, who analyzed the data, noted that because we don’t have the choice to give some people treatment and deny it to others in real life, “The game gave us a way to conduct an experiment on behavior that could never be done in real life.”
The experiment was conducted twice, each game taking place over 45 days. In one game, the cost for players to self-protect was low; in the other, it was higher. Players in the low-cost condition were significantly more likely to make the choice to protect themselves from infection.
About 50 players signed up for each game. Players were randomly assigned to different games. Each player’s result from each day provided one observation or data point.
Because self-protecting involves a cost – whether it’s time lost from work, the cost of burning gas while driving to get the shot or various other factors – players earned the highest number of points by staying healthy and not choosing the preventive measures.
That doesn’t mean the game rewards risk takers. “There’s a downside to taking risks,” Chen said. “If you get infected, you’re going to get lower points in the future. The game was set up so that if you’re healthy and you don’t do anything, you’re going to get 60 cents (or 0.6 points) a day. But if you get sick, you’re getting 10 cents per day.
“If you’re healthy but you take self-preventive action, it’s 35 or 45 cents per day. If you’re super safe, you’re going, ‘All right, I’m going to get my 35 or 45 cents a day.’ If you want the 60 cents per day, sure you can gamble and try to get that, but then that means you could get infected and end up with the 10 cents.”
Chen said the point system is designed to reflect the fact that “in our everyday lives, we have to take precautionary measures like washing our hands more frequently, going to get a flu shot, or in the context of STDs, having to wear condoms. These are things that are costly – maybe not necessarily in terms of money, but could be costly in terms of time, effort, loss of pleasure, etc.”
Chen looks forward to the next phase of the experiment, funding permitting.
“Because the framework is so flexible, essentially I get to play God in this game world,” he said. “We can change any aspect of the decision-making environment; I can change any aspect of the disease-transmission process.
“In the first set of experiments we ran, the disease has the property that if you get infected, you can recover, but if you recover you don’t have immunity – so you can get infected again. Where we want to go next is changing it so that we’re thinking more like something like the flu where you get infected, you could recover, and if you recover you can get some natural immunity.”
Given people’s many personality differences and various approaches to risk, Chen said he hopes the team’s findings ultimately show decision-makers that there shouldn’t be a one-size-fits-all approach to disease containment.
“If you try to come up with just one policy hoping that’s going to work for everybody, well, people have different attitudes, different psychological makeups, different experiences, different backgrounds,” he said. “So why would one policy work for everybody?
“I think perhaps a smarter approach would be to come up with not just one single policy but a set of policies where you’re taking into consideration these differences in some people and you’re tailoring the messages to different groups of people.”
Flu is “a pretty simple type of disease,” Griffith said. “The way people get sick, the way they recover, and the choices they’re making are fairly simple. The hope is to think about other setups and different ways of relaying the information and seeing how that changes the way people behave.”
Changed behavior by employers “is certainly the goal,” she said. “I think that if down the line we can create a substantial literature in looking at all these different settings that we’re interested in, that we can create some nice policy implications that if you keep costs down, if you provide the shots for free, if you do it at work so that all they have to do is get the shot on their lunch break … we’ll all see the benefits.”