Heuristics and Biases Approach

Thomas Gilovich and Kenneth Savitsky

As its name implies, the heuristics and biases approach to human judgment has both positive and negative agendas (Griffin, Gonzalez, & Varey, 2001). The positive agenda is to identify the mental operations that yield rapid and compelling solutions to a host of everyday judgmental problems. Most notably, Kahneman and Tversky identified a small number of automatic assessments – similarity, generation of examples, causal judgments – that are made rapidly in response to particular problems and thus exert considerable influence on the judgments that are ultimately rendered (Kahneman & Tversky, 1972; Tversky & Kahneman, 1974). When ascertaining the likelihood that someone is an engineer, for example, one cannot help but assess the similarity between the person in question and the prototypical engineer, and the resultant assessment is, at the very least, the starting point for the judgment of likelihood. Thus, the positive agenda is to understand what the processes of judgment are like.

The negative agenda is to understand what the processes of judgment are not like. Because assessments of similarity, the generation of examples, and causal judgments obey their own logic, everyday judgment will not always be fully “rational” and will not always conform to the laws of probability. Thus, Kahneman and Tversky demonstrated that people’s judgments are insufficiently sensitive to sample size, regression effects, prior odds, or, more generally, the reliability and diagnosticity of evidence. Their experiments were carefully crafted to reveal discrepancies between intuitive judgment and what is called for by the appropriate normative analysis.

Not surprisingly, the negative agenda appears to have garnered more attention and motivated more subsequent research – both hostile and congenial to Kahneman and Tversky’s original ideas – than the positive. This may be because accusations of irrationality raise the ire of some scholars and challenge the theoretical foundation of much work in the social sciences. In addition, a demonstration of what is wrong points the way to improvement – a prospect that may have particular appeal to psychologists, many of whom entered the discipline with some interest in improving the human condition. The focus on errors is also a natural outgrowth of the logic of hypothesis testing that constitutes the evidential core of the behavioral sciences (Kahneman, 1991). The laws of probability provide crisp null hypotheses to which actual judgment can be tellingly compared.

The logic of hypothesis testing and the negative agenda of the heuristics and biases approach have thus shaped much of the work on everyday judgment since the 1960s. Note, however, that this implies that judgments not so clearly at variance with a precise normative standard have received less attention, even if they are every bit as influenced by the same heuristic processes. In this article we wish to explore an important class of judgments that we believe have been relatively neglected for just this reason. In particular, we wish to explore the influence of representativeness on people’s assessment of causal relations – of what causes what.