someone flipping a coin had the same chance of correctly predicting whether a patient would die by suicide as an experienced psychiatrist — 50/50
This Psychologist Is Using A.I. to Predict Who Will Attempt Suicide
Diane Shipley
605

A number of things here. First, the CDC is the authority on death statistics. The number you quoted is right. This is just a note for future research.

Secondly, “die by suicide as a psychiatrist” implies people who are psychiatrists are killing themselves ;) I know what you’re trying to say here, but this entire sentence needs rewritten.

Third, if a statistical result ends in (roughly) 50/50 then this is a correlation error. It doesn’t mean people are “mediocre” in predicting suicide, it means that their current metric is 100% uncorrelated. In other words, they are SPECTACULARLY wrong in predicting suicide, not just mediocre.

If, for instance, psychologists were 80% wrong in predicting suicide, this would actually be BETTER because it implies that their measurements are working but their interpretation of the data is reversed. You could “bet against” the psychiatrist’s position and “win”. But with a 50/50 result, this shows that there’s literally no correlation and therefore the measurement itself is what’s flawed.

Show your support

Clapping shows how much you appreciated Sam Cel Roman’s story.