The mess behind the models: Too many of the COVID-19 models led policymakers astray. Here's how tomorrow's models will get it right

Abstract

If you wanted to “flatten the curve” in 2019, you might have been changing students' grades or stamping down a rug ripple. Today, that phrase refers only to the vital task of reducing the peak number of people concurrently infected with the COVID-19 virus. Beginning in early 2020, graphs depicting the expected number of infections spread through social networks, much like the virus itself. We've all become consumers of epidemiological models, the mathematical entities that spit out these ominous trend lines. Such models have existed for decades but have never received such widespread attention. They're informing public policy, financial planning, health care allocation, doomsday speculation, and Twitter hot takes. In the first quarter of 2020, government leaders were publicly parsing these computational speculations, making huge decisions about whether to shut down schools, businesses, and travel. Would an unchecked outbreak kill millions, or fizzle out? Which interventions would help the most? How sure could we be of any forecast? Models disagreed, and some people pointed to whichever curve best supported their predilections. It didn't help that the researchers building the models were still figuring out what the heck they were doing.