Solution
Praise and Punishment
Show the problem again
A flight instructor notices that cadets praised after excellent landings usually do worse on the next attempt, while cadets berated after terrible landings usually improve. He concludes punishment works and praise backfires. What statistical phenomenon better explains his observation?
Worked solution
Regression to the mean. Landing quality is skill plus noise. An exceptionally good landing likely involved lucky noise, so the next attempt tends to regress toward the cadet's average, appearing as decline; an exceptionally bad landing regresses upward, appearing as improvement. The pattern occurs with zero causal effect from feedback. This is Kahneman's real anecdote and the standard trap in evaluating interventions selected on extreme outcomes.
Source: Regression to the mean; the flight-instructor framing follows Kahneman and Tversky (1973). Statement written for AxiomIQ.