Base rate neglect is a term used in cognitive psychology and the decision sciences to explain how human reasoners, in making inferences about probability, often tend to ignore the background frequencies. For example, if the probability of any given woman having breast cancer is known to be 1/10,000, but a test on 10,000 women gives 100 positive results, reasoners will tend to overestimate the probability that any one of the women testing positive actually have cancer, rather than considering the possibility of false positives.
The analysis of base rate neglect is relatively recent within psychology, often considered a part of the heuristics and biases field. Rather than assuming that human beings are always rational thinkers, psychologists in this field explore the ways in which human judgements systematically deviate from the axioms of probability theory. These deviations occur because humans are often forced to make quick judgements based on scant information, and because the judgements which are most adaptive or rapid are not always the most correct. It appears that our species was not crafted by evolution to consistently produce mathematically accurate inferences based on a set of observed data.
The phenomenon of base rate neglect is also considered a part of descriptive decision theory, which studies how humans actually reason, as opposed to normative decision theory, which studies the best possible procedures for making any given decision. It has been found that human reasoners often ignore the base rate even when the information is easily available. This has yielded important results for the social sciences and economics, among other areas.
Base rate neglect is often mentioned in conjunction with Bayes' rule, a decision procedure which follows quickly from the axioms of probability theory. This rule demonstrates how to appropriately integrate base rates into new observations to provide updated probabilities in a consistent and accurate way. Therefore, deviation from base rates is also referred to as deviating from Bayes' rule.
Another example of base rate neglect in an experimental context would be the presentation to a group of test subjects of a list of ten students and descriptions of their habits and personalities. The presentation is followed by a question as to what grade point average any given student is likely to have. This information is presented together with base rate information on students' academic performance, which should guide test subjects in their guesses, but regularly does not. Given ten poor descriptions of students, test subjects will assign GPA estimates substantially out of line with base rates. Given ten positive descriptions, the bias occurs in the opposite direction. Presumably these skewed probability estimates occur every day in billions of human minds, with substantial implications for the way our society is operated.
By Michael Anissimov