[I wrote this general-audience piece during the last week of April and first few days of May. I pitched it to a few popular outlets, but couldn’t find it a home.]

What is the infection fatality rate for COVID-19? What groups of people are at greater risk, and what treatments are effective for lowering that risk? How many people have already been infected, and how many of them would be immune to reinfection? Which businesses are safe to reopen now, without triggering a second wave of infections, and under what guidelines? How can we conduct mass testing and contract tracing in a just and ethical way? What will the cultural and political fallout of this pandemic be for our society?

In the first days of May 2020, we know two things for certain about the answers to all of these questions. First, that almost nothing in certain. The data are incomplete, or unreliable, or simply won’t exist for months or years. Different models reach different conclusions, even when they use basically the same data and only subtly different methods. Preprints are available immediately, but haven’t been vetted by peer review, and may be retracted just as quickly. For any given study, it’s not hard to find a PhD on Twitter who has some scathing (and witty) criticism.

The other thing we know for certain is that we can’t wait until we’re certain. Policymakers at all levels — from Congress and FDA, through governors and state public health officials, down to mayors and ICU clinicians — need to make decisions now. Even inaction is a kind of action, as we’ve learned over the last three months from the contrasting cases of China and South Korea, New York and California. Any decision we make relies on answers to these questions, either implicit or explicitly.

A traditional view of the relationship between science and policy has two distinct steps. First scientists produce knowledge. Then policymakers pick up that knowledge and use it to make decisions. This view is simple and clean. Scientists have their job; policymakers have theirs. But it requires science to be finished before policymakers can act. We need an alternative, one that enables science to inform policy even as it tries to reduce uncertainty. Adaptive management provides exactly this alternative.

Adaptive management was originally developed in natural resources management and conservation biology. Ecosystems, such as forests and fisheries, are extremely complex systems. Organisms interact with each other and their environments in complex ways, and data are often indirect and patchy. Consequently, one forest may not react the same as any other forest in response to the same policy, and these differences can be all but impossible to predict in advance.

Adaptive management responds to this uncertainty by treat policies as science and science as policy. Like good experiments, good adaptive management policies compare interventions to control areas and systematically collect data. It’s important to identify the major outcomes of interest, while also being on the lookout for signs of unanticipated side-effects are “unknown unknowns.” Managers may even formulate precise hypotheses, and design policies that carefully test those hypotheses. Critically, as circumstances change and understanding improves, the knowledge produced by this scientific research is used to update and adapt policies. This requires ongoing, close collaboration between policymakers, researchers, and people on the ground in the area of interest, such as workers and local residents.

This does not mean that adaptive management is easy. It challenges deep-seated assumptions about both good policy and good science. On the policy side, stakeholders — often relevant businesses — push for “regulatory certainty,” meaning a stable policy landscape without economic surprises. In part for this reason, Fischman and Ruhl found that US federal agencies often practice “adaptive management lite,” which includes data collection but does not have any way for policies to adapt to improved scientific understanding. On the science side, some standards of “good science” — such as the slow process of peer review — might need to be adjusted or even sacrificed for the sake of rapidly informing policy. And the close collaboration with policymakers on controversial issues risks the accusation that science has been “politicized.” These challenges may explain why adaptive management has not been widely discussed outside of natural resources management policy. Nonetheless, DeFries and Nagendra include control of infectious disease as an area where adaptive management and related approaches can be especially valuable.

What would an adaptive management approach to lifting stay-at-home orders look like? First, data collection is essential. Communities must be able to know immediately when covid-19 appears in their area, and be able to respond rapidly using measures such as contact tracing. Second, policy changes might be patchy and staggered. Restrictions might be eased in one area a week or two before other areas. Some lessons learned in the early-open area might change the way opening is done elsewhere. Third, research ethics must be respected. Individual residents can’t give their consent to participating in policy experiments. But individuals can be informed about why policies are patchy, or might suddenly change. And individuals can also be given a chance to participate in making democratic policy-science. For example, town halls might be used to give residents an opportunity to share their experiences (data), interpret the most recent findings (science), and recommend courses of action (democracy). Similarly, policies must respect justice, recognizing for example that the burdens of the pandemic have fallen heavily on low-income communities of color.

Uncertainty does not need to be debilitating. But it does require thinking about policy as science and science as policy. And this, in turn, requires policymakers, researchers, and the public to work together.

COVID-19 needs adaptive management

Suggested Readings