Take probability models seriously, but not literally

Last week's Science had a feature article about several nurses who were convicted of murder in Europe based on statistical arguments regarding the number of babies who died under their care.  Some of these arguments appear to have been decisively debunked by statisticians not originally involved in the cases.  There is a sidebar article by the same author (Cathleen O'Grady) on mothers who were tried for the murders of their own children, again based on questionable statistical arguments.

These episodes are striking, given that I am currently reading a book by physicist Erica Thompson, Escape from Model Land:  How Mathematical Models Can Lead Us Astray and What We Can Do About It (Basic Books, 2022).  She argues that mathematical models always make simplifications and assumptions that divorce them from the real world; and while the modeling exercise may be of great instructive value, one must often be very careful not to claim that the model output represents reality.  In other words, once you step into "model land", and do your modeling exercise, you must not forget to step back out into the real world, and not mistake model land for reality.  The way I would put it is that models deserve to be taken seriously, but not literally.  In the case of the shoddy statistics that put innocent nurses and mothers in prison, those analyses should not be taken either seriously nor literally.