The prediction was chilling enough that it reportedly got President Trump to change course: An epidemiological model from Imperial College London showed that 2.2 million Americans could die if the government took no action to stop the spread of the coronavirus. But after the US and UK governments did take action, that model changed; it had previously predicted up to 500,000 deaths in the UK, but one of the scientists involved later issued a revised prediction of no more than 20,000 deaths. This was referred to as a "remarkable turn," a "retraction," and worse, but the truth, as Zeynep Tufekci explains at the Atlantic, is that the prediction changed because governments took action. "There was no turn, no walking back, not even a revision in the model," Tufekci writes in a column headlined, "Coronavirus Models Aren't Supposed to Be Right."
Rather, he explains, these models are meant to show a wide range of possible outcomes—and since epidemics grow exponentially, there is a vast difference between the most and least optimistic possible outcomes. In the case of the Imperial College model, those 2.2 million deaths represented the worst possible outcome. The job of the model is to convince authorities to ignore the most optimistic outcomes and even ignore the most likely, middle-range outcomes—in favor of focusing on the worst outcome and taking action to prevent it from happening. As we take action, "we are shaping the underlying parameters [of the model] themselves, because the parameters themselves are not fixed," he writes. "Sometimes, when we succeed ... it looks like we overreacted. A near miss can make a model look false. But that’s not always what happened. It just means we won. And that’s why we model." Read his full piece here. (Read more coronavirus stories.)