Prototype Heuristics and the Neglect of Extension

We offer a common account of three superficially dissimilar judgmental tasks: (1) categorical prediction (e.g., “In a set of 30 lawyers and 70 engineers, what is the probability that someone described as ‘charming, talkative, clever, and cynical’ is one of the lawyers?”); (2) summary evaluations of past events (e.g., “Overall, how aversive was it to be exposed for 30 minutes to your neighbor’s car alarm?”); and (3) economic valuations of public goods (e.g.,“What is the most you would be willing to pay to prevent 200,000 migrating birds from drowning in uncovered oil ponds?”). We propose that a generalization of the representativeness heuristic accounts for the remarkably similar biases that are observed in these diverse tasks.

The original analysis of categorical prediction by representativeness (Kahneman & Tversky 1973; Tversky & Kahneman, 1983) invoked two assumptions in which the word representative was used in different ways: (1) a prototype (a representative exemplar) is used to represent categories (e.g. bank tellers) in the prediction task; (2) the probability that the individual belongs to a category is judged by the degree to which the individual resembles (is representative of) the category stereotype. Thus, categorical prediction by representativeness involves two separate acts of substitution – the substitution of a prototypical exemplar for a category, and the substitution of the heuristic attribute of similarity for the target attribute of probability. Perhaps because they share a label, the two processes have not been distinguished in discussions of the representativeness heuristic. We separate them here by describing prototype heuristics, in which a prototype is substituted for its category, but in which representativeness is not necessarily the heuristic attribute.

The target attributes to which prototype heuristics are applied are extensional. An extensional attribute pertains to an aggregated property of a set or category for which an extension is specified – the probability that a set of lawyers includes Jack; the overall unpleasantness of a set of moments of hearing a car alarm; and the personal value of saving a certain number of birds from drowning in oil ponds. Normative judgments of extensional attributes are governed by a general principle of conditional adding, which dictates that each element of the set adds to the overall judgment an amount that depends on the elements already included. In simple cases, conditional adding is just regular adding – the total weight of a collection of chairs is the sum of their individual weights. In other cases, each element of the set contributes to the overall judgment, but the combination rule is not simple addition and is typically subadditive. For example, the economic value of protecting X birds should be increasing in X, but the value of saving 2,000 birds is for most people less than twice as large as the value of saving 1,000 birds.

The logic of categorical prediction entails that the probability of membership in a category should vary with its relative size, or base-rate. In prediction by representativeness, however, the representation of outcomes by prototypical exemplars effectively discards base-rates, because the prototype of a category (e.g., lawyers) contains no information about the size of its membership. Phenomena analogous to base-rate neglect are observed in other prototype heuristics: the monetary value attached to a public good is often insensitive to its scope and global evaluations of a temporally extended experience are often insensitive to its duration. These various instantiations of extension neglect (neglect of base rates, scope, and duration) have been discussed in separate literatures, but all can be explained by the two-part process that defines prototype heuristics: (1) a category is represented by a prototypical exemplar, and (2) a (nonextensional) property of the prototype is then used as a heuristic attribute to evaluate an extensional target attribute of the category. Extension neglect in all its forms is most likely to be observed in between-subjects experiments. Within-subject factorial designs consistently yield the additive extension effect: judgments change when people evaluate different amounts within the same experiment. The source illustrates this in Fig. 2.3, which is not reproduced here.