Contributed by Ronald Baecker, who is an Emeritus Professor of Computer Science at the University of Toronto, author of Computers and Society: Modern Perspectives, co-author of thecovidguide.com, and author of the forthcoming Digital Dreams Have Become Nightmares: What We Must Do.
A forecasting model is a prediction of how the world will evolve, of what will happen in the future with respect to some phenomena (such as the motion of objects, the financial health of a business) or the spread of an epidemic.
Models are sometimes expressed as equations. Newton’s Second Law of Motion, often stated as F = M × A, describes the relationship between the force F acting on an object, its mass M, and its acceleration A. It is useful. For example, it can predict how quickly a hockey puck weighing M ounces will accelerate towards the net if it is struck by the hockey stick of a player whose strength will result in a certain force applied to the puck.
Models can also be expressed as computer programs, and in spreadsheets which allow assumptions to be expressed without writing a program. These also are useful. For example, wise modern business owners build spreadsheets forecasting their profit-and-loss, cash flow, and balance sheets. If the assumptions build into the model are good ones, the spreadsheet will help them know when they might need a loan, or when they might reach profitability.
The word “might” here is key. Newton’s Laws have been validated in countless ways over centuries. They are guaranteed to hold true in all normal situations that we encounter as humans, even though they do not hold true at a very tiny scale, a world understood by quantum mechanics, or a very large scale, a world understood by theories of relativity. Such certainty is not the case with a typical spreadsheet built by a business owner, who makes assumptions in the model that may or may not be valid.
Recently, models have become political weapons, to be used when convenient, to be ignored or hidden when not convenient. For example. the American Civil Liberties Union (ACLU) in Idaho questioned the use of an Excel spreadsheet that was being used to justify cutting the level of Medicaid assistance given to individuals with developmental and intellectual disabilities. When asked about the logic embedded in the spreadsheet, Medicaid refused to disclose it, claiming it was a ‘trade secret’. A court granted an injunction against the cuts, and ordered the formula made public. It soon became clear that the spreadsheet had numerous errors in it. A 2015 class action suit against the state of Idaho is still being deliberated by the Idaho Supreme Court (for the second time).
Pandemic forecasting models guide the life-and-death decisions about how quickly physical distancing rules or guidelines should be relaxed by various jurisdictions. They also have become weapons. A recent case occurred in the US state of Arizona. On May 5, Donald Trump visited Arizona. On May 6, hours after the State’s governor relaxed Stay at Home restrictions, the Arizona Department of Health Services shut down a project in which approximately ten University of Arizona researchers had developing an Arizona-specific model to guide public policy with respect to such restrictions. One rationale given was that the model was not needed, because the U.S. federal agency FEMA (Federal Emergency Management Agency) had its own model. That decision was rescinded two days later after a public outcry.
The problem is that the FEMA model and the algorithm that determines its predictions are secret. Nobody from the media, from the medical establishment, or the public can examine the assumptions used to derive conclusions and to guide policy, and to help it be wise policy. This is dangerous — a good example of science and mathematics being used for political ends. See here, here, and here to learn more about this topic.
FOR THINKING AND DISCUSSION
How should society engage with the creators of pandemic models and the assumptions that animate them?