Cause Trumps Statisitics gives examples of statitistical effects related to false positives in medical tests, although that is not an example Kahneman used. He does mention that eyewitness testimony, even when probably in error, will overwhelm statistical probabilities in most people’s minds. He uses an example about the color of a cab involved in an accident where one color cab is much more prevalent and an eyewitness is expected to be right 80% of the time. The situation is mathematically the same as the false positive scenario in medical tests.
Even dramatic cases and experiments where an alleged participant feigned choking and asked for help (but got less than expected because everyone assumed others were helping) did not overcome this bias in favor of the story over the statistics. Kahneman does describe a way to turn this affect to your benefit by telling the story of people who display the most common behavior.
Chapter 17 is Regression to the Mean. I knew all about this effect. What I had not thought about which I found interesting was the influence this has on trainers. When a trainer gives positive feedback it is usually because a trainee has done something spectacular. So they are likely to do worse next time just due to luck. On the other hand if a trainer yells, or criticizes severely it is usually because the trainee did particularly poorly. Regression to the mean says the trainee is unlikely to do so poorly again. So, from the point of view of the trainer the feedback they receive is that yelling and generally behaving boorishly improves outcomes while positive reinforcement causes worse outcomes. If a trainer does not think of regression to the mean and ignore the brain’s desire to assign everything a cause they can develop quite a misleading perception of the best way to teach.
Because I knew about both of the major effects before I read this I classify this in “Just a Thought,” but include them for completeness on my writing about Thinking, Fast and Slow.