No COVID-19 Models Are Perfect, But Some Are Useful

Published: May 2020

Authors: Peter Gleick

Pages: NA

 

Op-Ed

The global coronavirus pandemic has brought renewed interest and focus on scientific models as we try to get a handle on what the future will bring, how many people will fall sick and die, what the economic impacts will be, and what actions politicians should take. But confusion abounds about what these “models” say and how to reconcile their often seemingly conflicting visions of the future. Recent political attacks on these models reflect a lack of understanding about what models are, how they work, and their usefulness and limitations. Conservative Fox News commentator Laura Ingraham attacked models on her show. Senator John Cornyn, a Republican from Texas tweeted “After #COVID-19 crisis passes, could we have a good faith discussion about the uses and abuses of ‘modeling’ to predict the future?”

Models are all around us. Without knowing it, we all use models all the time to try to understand outcomes of complex situations. The decisions you make on how to spend your monthly paycheck or save for retirement are financial models. The car you drive and the toaster in your kitchen were both designed with engineering models. Advertisers make models of consumer behavior, preference, and consumption of media when they design and buy ads. These models depend on science, but also on human behavior and actions that are far less predictable.

Even a recipe for bran muffins is a model—and a good example of the kind of models we use every day without thinking about it. Cooks combine centuries of knowledge about the chemical behavior of different ingredients with their personal experiences to create a model—a recipe—for what they hope is a delicious bran muffin. But whether the “bran muffin model” actually produces a good muffin or a burned hockey puck depends not just on the recipe but on factors completely out of the control of the recipe designer. When you set your oven for 400°F, does it heat to only 350°? Will you mistake a teaspoon of salt for a tablespoon of baking soda? Will you fall asleep and burn the muffins? Will the recipe maker’s tastes match your own?

Because of these uncertainties and unknowns, scientists who work with models try not to call the outcomes “predictions”—rather we call them “projections” or “scenarios.” A prediction implies more accuracy and certainty than many models provide. For all these reasons, scientists often repeat the classic aphorism “All models are wrong, but some are useful,” by which we mean models are only as good as our understanding of the scientific knowledge that goes into them. But useful models help us understand how science and human choices interact, providing valuable insight for policymakers.

This op-ed originally appeared in TIME.