A review by michael5000
Noise: A Flaw in Human Judgment by Cass R. Sunstein, Daniel Kahneman, Olivier Sibony

3.0

A good book, but not a good Kahneman book. The cognitive "noise" that the book talks about is real, certainly, and much of the data that they present supports that finding. But they are remarkably slow to notice that much of the noise in their data comes from a mismatch between what judgements are nominally about, and what judgements are REALLY about. For example, when they say that algorithms are stronger than individual evaluations at predicting future employee success, they seem only dimly aware that a person filling out an employee evaluation is not necessarily trying to predict future employee success, and is almost certainly not EXCLUSIVELY trying to predict future employee success. In cases like this, an algorithm approach isn't trumping human judgement by reducing cognitive noise; it's just forcing judgement into a single dimension by eliminating the meta-game. Kahneman & co. sometimes indicate that general idea has occurred to them, but for a lot of the research under discussion they think it is the sideshow when it is pretty obviously the main event. Which is a shame, both because it weakens the book, and because they would have been able to articulate what I'm talking about SO MUCH BETTER than I'm managing here!