If you take a close look at your local or state-level data on the incidence of coronavirus, you might see an interesting up-and-down pattern. Notice the spikes and valleys marked with arrows on the graph of my county's data above: Cases don’t simply grow or decrease, they wobble around. Public health experts use 3- or 7-day moving averages to smooth out these bumps, but you can see peaks and dips in the fitted lines as well. Sometimes one of the spikes leads to exponential growth, as happened in many places in June, and may take months to come back down again. Unfortunately, we probably won’t know that the next increase is more than a temporary increase until it’s too late to change things.
The cyclical ups and downs suggest that something is wrong with our approach to the COVID-19 pandemic. I believe that this comes out of a basic property of the human risk-perception system, which leads us to be reactive rather than proactive in our health decision-making. We give too much weight to what has just happened as we adapt to changing conditions, and we are insufficiently attentive to what’s potentially coming next. There are plenty of examples of reactive decision-making throughout healthcare: For instance, think about the person with high blood pressure who doesn’t have symptoms (predominance of reactive decision-making), and so goes on about his life with little concern for long-term risks of heart attack or stroke (absence of proactive decision-making).
In the context of coronavirus, reactive decision-making might look like resuming on-campus college classes because there have been few outbreaks at colleges (of course, that's at least partly because they were closed), and then quickly bringing students home when an outbreak does occur (despite the fact that this could spread the virus back to one’s home community). Or in the case of masks, people seem to wear them more regularly when infection rates go up, and then become more lax at times or in areas with lower infection rates (even though cloth masks are useful mainly for preventing spread; when rates are high people should just stay home). There is even some suggestion that people are actively monitoring public health data and then increasing risky activities like travel or restaurant dining when incidence rates go down (which of course potentially fuels the next spike in case numbers).
I have previously argued that the difference between long-term and short-term thinking is often a function of whether the Narrative or the Intuitive mind is engaged when thinking about a problem. But coronavirus prevention isn’t a problem of intention-behavior gaps for many people; instead, it’s our intentions that need work. People who aren't wearing masks or maintaining social distance from others are also more likely to argue that masks are ineffective, that children can't be infected, or that only 6% of otherwise young and healthy people die from coronavirus alone. (Note that the first two of these are false; the last one is true but is cold comfort to the estimated 50% of Americans who live with one or more chronic diseases). Our inconsistent coronavirus response is thus fundamentally a problem of the Narrative mind, although some Intuitive features of perception may be contributing to our difficulty in thinking accurately about the risks.
The big problem in making future-oriented decisions is, of course, uncertainty. The following graph from Kahneman and Tversky’s Nobel-prize-winning prospect theory shows how people make decisions when the future is uncertain. In the upper left-hand part of the graph, the curve twists sharply upward to show that when perceived benefits from some action are great (the x-axis), we over-estimate its likelihood (the y-axis). In a perfectly rational world, the line would be straight and at a 45 degree angle (in other words, y = x). But the degree to which we want something distorts our judgment, and makes us think that what we want is what we will get even when the chances aren't so good.
On the lower right a slightly different pattern occurs, in which we over-inflate the likelihood of risks when we focus on potential losses. The more terrible the potential loss, the more likely we estimate it to be — in other words, we are risk-averse. Notice that the curve goes down on the left at a sharper angle than it goes up on the right: The scarier the potential consequences of some event, like coronavirus infection, the stronger our desire to avoid it, and the more we are willing to incur costs to do so. This is probably why initial “flatten the curve” efforts in March and April were successful, when so much about the novel coronavirus was still unknown =and frightening case examples dominated the news.
Over time the potential risks associated with coronavirus infection started to seem less serious: Most of us still hadn’t been infected, after all, and neither had most people we knew. News stories about treatment advances and the high rate of asymptomatic infection made the disease seem statistically less dangerous. New data in May made it seem that touch-based transmission was unlikely (although now it looks like that was also too simplistic a view), and that we could keep everyone safe just by wearing masks and staying 6 feet apart. Finally, by June the new infection rate had finally started to go down and people felt safer. People were then primed to focus on the gains of re-opening (over-weighting them relative to their actual utility), and to under-weight potential losses. This primed us for a drastic resurgence of coronavirus cases in June and July.
Unfortunately as our risk-benefit equation changed our focus also shifted. Beyond the general change in thinking produced by a focus on benefits instead of risks, we also fell victim to confirmation bias in terms of the new information that we were paying attention to. We started to pay less attention to new findings, like the risk of “long haul” COVID-19 symptoms in about a third of people, the potential cardiac or neurological consequences of even mild infection, the new evidence suggesting that contaminated surfaces are the primary cause of infection in 20-30% of COVID-19 cases, and the mounting evidence that SARS-CoV-2 can linger in closed air spaces and travel much farther than 6 feet. People took more risks, and case numbers spiked.
Because the virus is so hard to control, even with draconian shutdown methods, it took a very long time for numbers to come down once again. Why did they? The simple explanation is mask mandates, although many people still disobey them. The more likely cause of the decrease is that people saw the numbers going up, that trend re-focused them on the risk of infection, and they were more likely to do the one thing that’s empirically proven to reduce viral spread — they stayed home. In Colorado, the level of social distancing was 80% in May, but less than 50% in June based on public health department estimates. By August the level of distancing had risen back to 75%. Now that case numbers have finally come down once more, people might again decrease their level of social distancing and another surge will follow. Also notice that the current plateau is higher than the one in June: Another problem is that people simply get tired when they hear about constant threats, so we are gradually losing ground.
As the intensity of a stimulus increases (x-axis) the frequency with which the neuron fires also increases (y-axis), but notice that the relationship is again not one-to-one. Up to a certain level of input from the environment the neuron fires infrequently, but after that its activity jumps up dramatically and out of proportion to the stimulus level. This on-or-off neural firing pattern mimics the pattern of behavior shown by prospect theory.
How does this inform public health interventions to reduce COVID-19 risk? First, we need to recognize that most people are not going to manage an "average" level of risk over a sustained period of time; they are instead likely to take risks in bursts. Second, recognizing this pattern, it might be better to have only some groups of people taking risks at any given point in time. For example, we could assign a preferred day of the week for shopping based on last name, to encourage a steady flow of shoppers throughout the week. Or we could use apps to manage traffic flow in stores or other high-volume locations. These strategies are likely to be unpopular because they use external controls to manage behavior, but they could overcome the difficulty that people have in managing risks successfully on their own. Third, we could continue to publicize COVID-19 risks including some that were not known early in the pandemic, and continue to emphasize secondary benefits of staying home (e.g., less air pollution, less commuting stress), rather than emphasizing the problems that make physical distancing harder to tolerate. Finally, we could explore simulated or carefully managed real-life experiences that show people the benefits they imagine are not as good as they seem. In a trivial example, my family cooked all our own food for 11 weeks at the start of the pandemic and I found myself craving french fries; once we finally got take-out I realized that they weren't actually as good as I remembered. Giving people a small and relatively safe taste of what they are missing might help them to tolerate its absence better than simply lecturing them on why they can't have it.
Comments
Post a Comment