We are thinking and talking about the COVID-19 models all wrong, writes Kevin Schulman on The Agitator, a normally smart blog on non-profit fundraising. “Most believe these models represent attempts at prediction rather than as a managerial, scenario planning tool for which they are intended,” he argues — as if believing a model is a “managerial, scenario planning tool” would make it any more accessible.
Zeynep Tufekci (yes, I’m writing about her a lot lately) makes much the same case in The Atlantic. “The problem is sometimes that people believe epidemiologists, and then get mad when their models aren’t crystal balls,” she writes. Well, yes: when you shut down the global economy and risk worldwide depression based on modeling, some people might get mad when the model changes based on new inputs or revised assumptions. Epidemiology should now come with hazard pay.
The real question is not whether we are thinking all wrong about models, but whether talking differently about models could get us to think differently about their purposes and how they should be influencing our actions. (Although, quite frankly, the now-famous Imperial College report presented four scenarios, and at least two scared the bejeezus out of the Boris Johnson cabinet and the Trump Administration. Now that was a scenario planning tool.)
The model that’s taken over in the public eye — the University of Washington’s Institute for Health Metrics and Evaluation (IHME) COVID-19 model — updates real-time and projection curves daily for the United States and each U.S. state for cases, medical resources needed to deal with the cases, resource shortages, deaths per day and total deaths. Governments and health care administrators are using the IHME model for granular planning and responses. For most of the rest of us nonspecialists, this model is now simply a grim scoreboard, and it’s a bit elitist to ask us to respond to it otherwise. We might be bending it, but its curve is still a wave, about to crash upon us:
But are there other ways to depict and communicate these models that could help us see ourselves as actors — not passive home-isolators? That could promote social distancing even when we get on the backside of these curves, when that distancing is still necessary but will be oh-so-tempting to stop?
The best part of Tufekci’s Atlantic piece is when she talks about COVID-19 models as trees with branches, with some of those branches representing the worst-case scenarios:
With COVID-19 models, we have one simple, urgent goal: to ignore all the optimistic branches and that thick trunk in the middle representing the most likely outcomes. Instead, we need to focus on the branches representing the worst outcomes, and prune them with all our might. Social isolation reduces transmission, and slows the spread of the disease. In doing so, it chops off branches that represent some of the worst futures. Contact tracing catches people before they infect others, pruning more branches that represent unchecked catastrophes.
At the beginning of a pandemic, we have the disadvantage of higher uncertainty, but the advantage of being early: The costs of our actions are lower because the disease is less widespread. As we prune the tree of the terrible, unthinkable branches, we are not just choosing a path; we are shaping the underlying parameters themselves, because the parameters themselves are not fixed. If our hospitals are not overrun, we will have fewer deaths and thus a lower fatality rate. That’s why we shouldn’t get bogged down in litigating a model’s numbers. Instead we should focus on the parameters we can change, and change them.
The tree metaphor could make immediately clear what all of us — policymakers, corporate leaders, ordinary citizens — have to continue to do. We choose the path every day; we cut the “terrible, unthinkable branches” every day.
I am amazed, given how confusing and misleading and uninspiring much of our government’s communications have been throughout this crisis, at how much of the United States is still working to bend the curve — a curve that represents not just death and suffering, but also the sum of our collective action to staunch those casualties.
But a curve’s collectivity also signals on a foundational level a kind of unfathomable mass. Like the curves climate change activists use, the COVID-19 curves dwarf individuals and their actions. As such, they are fundamentally discouraging of that action.
So I wonder how much earlier we would have acted had we also had Tufekci’s tree metaphor and were able to visualize as broadly as the curve that needed flattening. Could we have gotten more people chopping off the worst branches of the tree? Could we have prompted them to cut off their ability to infect thousands of others, as in this New York Times graphic from a few weeks back?
Models are metaphors, just like any other form of research communication. We’ve been depicting our models as impersonal, collective waves — which is a short step from victimization. But it’s not too late to add a tree.