Book review: Not Bayesian enough. Rationality in an Uncertain World: Essays on the Cognitive Science of Human Reasoning. Mike Oaksford and Nick Chater. Psychology Press, Hove, Sussex, 1998. No. of pages 336. ISBN 0-86377-534-9
✍ Scribed by David Over
- Publisher
- John Wiley and Sons
- Year
- 2000
- Tongue
- English
- Weight
- 51 KB
- Volume
- 14
- Category
- Article
- ISSN
- 0888-4080
No coin nor oath required. For personal study only.
✦ Synopsis
Almost all the papers in this collection have been published before, though some have been adapted to some extent, perhaps with some bene®t of hindsight, and the introductory and concluding chapters, 1 and 16, are mostly new. Oaksford and Chater's work is always extremely stimulating and some of the papers collected here have rightly had a big impact, with many people greatly impressed by them and some others strongly reacting against them.
The essays in Part 1 are a vigorous attack on what Oaksford and Chater call logicist cognitive science, which is the view that ordinary people's reasoning is the result of the operation of some kind of mental logical system. No one has ever claimed, of course, that this mental system could exactly follow a perfect logic as axiomatized by expert logicians, but many cognitive psychologists have held that some axiomatization of logic could be modi®ed to account for people's limited and sometimes ¯awed reasoning. Oaksford and Chater bring out very well the problems with present theories of this type, whether they postulate what is usually called a mental logic or alternatively the construction of mental models. A mental logic is a supposed mental system based on a modi®cation of a natural deduction logic, consisting of introduction and elimination inference rules for the logical connectives, such as the elimination rule of Modus Ponens for the conditional. The alternative construction of mental models is supposed to generate representations which have a direct interpretation in logical semantics, like disjunctive normal forms in propositional logic or sets of mental tokens for predicate logic. Oaksford and Chater's most basic objection to these uses of logic in psychology is that they cover little of ordinary reasoning, which has to take account of all the sources of uncertainty there are in the real world. Present mental logics are wholly inadequate in the face of uncertainty, and the great proponent of mental models, Johnson-Laird, has only recently started to extend his theory to uncertain reasoning.
Part II of the book contains Oaksford and Chater's papers which make a much more positive and original contribution to the study of reasoning. They hold that logic should be replaced by probability theory as the basis of cognitive science and psychology. Deeply in¯uenced by Anderson's programme of rational analysis, they propose to use Bayesian probability theory to give such analyses of problems in reasoning and decision making. This is best illustrated by their papers in the book on the selection task, which have stimulated much comment and further work. The psychologists who ®rst studied that task thought that people are irrational because, when investigating a conditional of the form `if p then q', they tend not to select a not-q card which might have a p on the other side. In other words, people appear to ignore a possibility which might falsify the conditional, and they seem to have a con®rmation bias or some other supposedly irrational or non-rational bias. But as Oaksford and Chater point out, it is grossly inecient to investigate many ordinary conditionals by examining not-q instances to see if these are also p instances. The best example of this is the classic one from the philosophy of science, stating that, if something is a raven, then it is black. No rational person would try to ®nd out whether this is true by examining non-black things to see if any of these are ravens. There are too many non-black things and too few ravens in the world for this to be rational. More formally, this notion of rationality can be clari®ed in Bayesian probability and con®rmation theory. Whether or not all ravens are black, it is practically certain that