Interview with Philip Johnson-Laird in The Reasoner
Hykel Hosni recently interviewed Phil Johnson-Laird in the latest issue of The Reasoner. Here’s a short excerpt:
HH: How about probabilistic reasoning?
PJL: The idea that probabilities enter into reasoning is quite popular at the moment: theorists want to replace logic with the probability calculus to account for reasoning. I’m skeptical. People are unlikely to adduce probabilities in reasoning unless the task itself suggests that they do. Two pilots asked me to adjudicate a disagreement between them. They were arguing about the likelihood that both engines of a twin- engined plane failed. The pilot who flew jets in Vietnam said: double the probability of one engine failing, whereas the light- plane pilot said: halve it. In a spirit of compromise, I told them that they were both wrong. What this anecdote confirms is that people don’t know how to calculate the probability of a con- junction of two events. The model theory predicts that when the two probabilities differ they take the average. Sunny Khemlani and Max Lotstein corroborated this procedure, and similar ones for disjunctions and conditional probabilities. We are all duffers about probabilities until we have mastered the calculus, and most of us remain so afterwards. You’re not a duffer, if you can answer this question: what three probabilities fix the complete joint probability distribution for two events, and no matter what numerical values you guess for them, its sum is always 100%?