Epistemology and the Psychology of Human Judgment

by Michael A. Bishop, J.D. Trout

Bishop and Trout here present a unique and provocative new approach to epistemology (the theory of human knowledge and reasoning). Their approach aims to liberate epistemology from the scholastic debates of standard analytic epistemology, and treat it as a branch of the philosophy of science. The approach is novel in its use of cost-benefit analysis to guide people facing real reasoning problems and in its framework for resolving normative disputes in psychology. Based on empirical data, Bishop and Trout show how people can improve their reasoning by relying on Statistical Prediction Rules (SPRs). They then develop and articulate the positive core of the book. Their view, Strategic Reliabilism, claims that epistemic excellence consists in the efficient allocation of cognitive resources to reliable reasoning strategies, applied to significant problems. The last third of the book develops the implications of this view for standard analytic epistemology; for resolving normative
disputes in psychology; and for offering practical, concrete advice on how this theory can improve real people's reasoning.

This is a truly distinctive and controversial work that spans many disciplines and will speak to an unusually diverse group, including people in epistemology, philosophy of science, decision theory, cognitive and clinical psychology, and ethics and public policy.

Genres: Philosophy, Psychology, Nonfiction

224 pages, Paperback

First published December 1, 2004

Reviews

[Bria · Rating 3 out of 5]

You know which audience they're writing for when they take the time to explain a passing use of the phrase "jump the shark" but never define the specific philosophical terms they spend pages debating about.

[Soroush · Rating 3 out of 5]

Read this book, only if you have an interest in naturalized epistemology. On the one hand, if you're not interested in epistemology or don't know what it is, then this book will be quite dry and dense for you. On the other hand, if you are interested in epistemology and have studied other books and papers on epistemology before, then you may find this book unfulfilling. Why? Because Bishop and Trout take a radically different approach to epistemology without adequately justifying their foundations. Nonetheless, their theory, Strategic Reliabilism, is an interesting theory. If you're not concerned about the problems surrounding their Aristotelian Principle (the axial pillar of their theory) or their position and flip flopping on the topic of intuition, then you may enjoy the mechanism underlying their approach to epistemology. In short, on the surface their theory seems like an adequate substitute for the old approach to epistemology. However, at the root of their theory, there are serious problems which put them along side the traditional analytic approach to epistemology.

[Gus Lackner · Rating 1 out of 5]

This book contains a reasonable criticism of what it calls the "Standard Analytic Epistemology," but it fails to provide a cohesive alternative. "When you're outnumbered," write Bishop and Trout, "and you want to show your theory is possible, proposing an actual theory is best" (p.23). Their positive content boils down to this: pro cons lists are an easy way to make better decisions. They couch this in overly complicated language about cognitive resources, normative applications, orders, cardinality, measurable irrationality, etc. They use the word "epistemology" just about once a sentence and still have to gall to write: "[When considering] an environment full of objects that are phenomenologically identical but ontologically distinct, SAE jumped the shark" (p. 22). Phrases like this are important and do have meaning. This one, for example, allows philosophers to formally engage with a general form of the problem 'all [race] people look the same.'

Trout and Bishop's most egregious errors are in their statistical thinking. For example, let us look at their discussions on pages 34 and 130. Bishop and Trout discuss Condorcet's jury on page 34, "if a jury is facing a binary choice and each jury member makes her decision independently and has a better-than-even chance of making the right decision, this will tend toward certainty as the number of jurors tends to infinity. We can think of the successful linear models we have introduced as a jury" (emphasis mine) just after listing the axiom "the evidential cues must be somewhat redundant. For example, people with higher GPAs tend to have higher test scores." To anyone with basic statistical literacy, it is clear that independence and endogenous correlation cannot coexist.

Bishop and Trout discuss a puzzle on page 130. While not in error, the discussion lacks a huge amount of context. It is a featherless biped moment; despite a strictly correct solution to the puzzle, I believe Bishop and Trout induct too far and that further research is necessary. They conclude from a specific connection of two binary choices that people are incapable of logical thinking because the proper action, given the highly contrived situation, is to update the first probability with the second. In arguing for a Bayesian thought process (epistemology), they overlook the heuristics which rely on correlation that people bring to the situation. In this specific problem, they decide that a test should not be administered because it would lead to a net loss of life since it does not update away demographic differences. If the test were at all correlated with the unspecified 'demographic reasons' this is an invalid recommendation. If the test's success correlated with something easily measurable in addition to the test, this is an invalid recommendation; a smart doctor would simply check for the thing that correlates with the test working and decide whether to administer the test. In real life, most outcomes are continuous, not binary. For this reason, people may develop epistemologies which are based on the non-ergodicity of stochastic processes rather than the frequentist view implied in this problem. Just as people naturally perceive count logarithmically but learn to count linearly, it is likely that people perceive processes in a non-ergodic stochastic manner and learn frequentist thinking. It is not clear that the latter is more strictly "rational" than the former.

Despite my gripes, I fête Bishop and Trout's work of translating the basics of statistical thinking into the language of philosophy for the recruitment of more minds to the cause.

[Boris · Rating 5 out of 5]

The best book in philosophy on the topic of epistemology. Simply a must read for any philosopher.

[Justin]

He's right about just about everything but yet misses so much.