The Drunkard's Walk

Mlodinow, a physicist, wrote an accessible book on randomness and probability, combining the history of the mathematics with evidence that people reason about it badly.

The historical thread runs from Cardano's early work on gambling odds through Pascal and Fermat's correspondence, Bernoulli's law of large numbers, Bayes and the problem of inverse probability, Laplace, Gauss and the normal distribution, and the development of statistics. Mlodinow explains each concept through the problem its originator was trying to solve.

The psychological thread reports the standard findings on probabilistic reasoning: the conjunction fallacy, where a specific conjunction is judged more probable than one of its components; base rate neglect, illustrated with medical testing where a positive result on an accurate test for a rare condition usually indicates no condition; the availability heuristic; the misunderstanding of conditional probability shown by the Monty Hall problem and by the prosecutor's fallacy; and the tendency to find patterns in random sequences, including the clustering illusion and the hot hand.

The applied argument concerns how much of observed success is randomness. Mlodinow examines mutual fund manager records, corporate executive performance, film studio executives, wine ratings, publishing decisions and sports streaks, arguing in each case that the observed distribution of outcomes is consistent with chance operating on small underlying differences, and that our tendency to construct causal narratives after the fact makes randomness invisible.

The book covers ground also covered by Taleb and Kahneman. Some of the psychology studies cited come from literatures affected by replication problems since 2008, though the core probability findings have held. The treatment is introductory and a reader with statistical training will find little new.