Tuesday, March 15, 2011

The Drunkard's Walk

With my usual degree of timeliness I read The Drunkward's Walk by Leonard Mlodinow. It begins with the wonderful anecdote of the Spanish man who won the lotto with a number ending in 48 and explained that he dreamt of the number 7 for 7 nights and since 7 times 7 is 48 then it was clear to him what to do!

Drawing from the history of probability theory and statistics, the book address how people form judgments and make choices under uncertainty. It has 10 chapters. Chapter 1 "Peering through the eyes of randomness" is a beautiful discussion of the role of chance in our lives and the limits to our use of intuition to understand this role. Includes, among a lot of other things, the interesting factoid that the first Harry Potter novel was rejected by nine publishers! Chapter 2 "The Law of Truths and Half-Truths" outlines some basic probability rules and how they can sometimes confound intuition. Includes a discussion of the representativeness heurisic. Former students will recognise Linda! Interesting discussion of the greek deficiency in probability. Chapter 3 "Finding your way through a space of probabilities" goes through subtle rules of probability. I still feel stupid when I see the Monty Hall problem solution, having seen it first in the late 90s though I enjoyed Mlodinow's account of it a great deal particularly as it shows that even people like Paul Erdos literally had to have simulations shown to him before he would accept the answer. Mlondinow knits the biography of forgotten Italian statistician Cardano into an account of how to solve Monty-Hall type problems through representing them as state-spaces. Chapter 4 "Tracking the Path to Success" outlines Pascal's contribution to probability theory. Chapter 5 "The Dueling Law of Large and Small Numbers" deals with random numbers and probabilistic expectation. Among many great pieces, includes the anecdote I hadn't heard that Pascal is rumoured to have invented the roulette wheel while playing with an idea for a perpetual motion machine. Chapter 6 False Positives and Positive Fallacies examines Bayes rule. Chapter 7 "Measurement and the Law of Errors", among other things, outlines normality and statistical distributions. Chapter 8 "Order in Chaos" examines statistical regularity. Chapter 9 "Illusions of Patterns and Patterns of Illusions" examines how people impose flawed patterns on data e.g. illusory cancer clusters, baseball hothands, luftwaffe bombing clusters. Chapter 10 "Drunkwards Walk" further examines the role of flawed expectations and misconceptions of chance.

My favourite aspect of the book is the rich and complex anecdotes and examples he uses to illustrate the main concepts. I like the fact that in a discussion of medieval preference for incantations over analysis, you suddenly get a glimpse of a modern bond trader explaining to a CNN reporter that they worry about using the "wrong" toilet before a major trade. The book sets the judgement and biases programme in a history of statistics narrative and this creates a number of great points of contact between modern judgement research and the lives of people like Cardano, Galileo and Pascal.

I don't know if the publishers will see this blogpost but this book would be a brilliant RSA-animate production.

No comments: