Recognition Heuristic

Heuristics: Deliberate or Automatic?

So far, we have described judgment by heuristic as an intuitive and unintentional process of attribute substitution, which we attribute to System 1. However, attribute substitution can also be a deliberate System 2 strategy, as when a voter decides to evaluate candidates solely by their stance on a particular issue. In other cases, a heuristic is both initiated spontaneously by System 1 and deliberately adopted by System 2. The recognition heuristic proposed by Gigerenzer and his colleagues appears to fall in that class.

Experiments described by Gigerenzer and Goldstein (1996; see also Gigerenzer et al., 1999) show that respondents rely on feelings of familiarity and unfamiliarity to compare uncertain quantities, such as the relative size of two cities. For example, 78% of a sample of German students recognized San Diego as an American city, but only 4% recognized San Antonio, and every student who recognized San Diego but not San Antonio concluded (correctly) that San Diego is larger. Though far from perfect (the correlation between actual population and recognition was only 0.60 in that experiment), the recognition heuristic is surely a reasonable strategy for that task. Indeed, when students were given pairs of the 22 most populous cities in the United States or Germany, Americans slightly outperformed Germans when comparing the size of German cities, and Germans did slightly better than Americans when judging American cities (Gigerenzer & Goldstein, 1996).

Gigerenzer and his colleagues have described the recognition heuristic as a deliberate strategy, which in our terms is an operation of System 2. This description seems highly plausible. In addition, however, we have proposed that familiarity is an attribute that System 1 evaluates routinely, regardless of the current judgment goal. On this view, the recognition heuristic has an automatic component, which could be studied by varying tasks and by measuring reaction times. Imagine a reaction-time study in which respondents on each trial see a question such as “Which city name is printed in larger font?” or “Which city name contains more vowels?” immediately followed by a pair of cities that differ in familiarity. Research on conceptual Stroop effects (e.g., Keysar, 1989) suggests that the more familiar city name will be the favored answer to any question that is associatively related to prominence, size or quantity. On this hypothesis, errors will be systematic, and response times will be faster for compatible than for incompatible responses. An even more radical possibility, arising from the work of Gilbert (1989), Begg (see. e.g., Begg & Armour, 1991; Begg, Anas, & Farinacci, 1992), and Mandler (see, e.g., Mandler, Hamson, & Dorfman, 1990) is that there will be a bias favoring the familiar item as an answer to any question – perhaps even “Which city is smaller?” or “Which city has fewer dentists?” If either of these hypotheses is correct, the recognition heuristic belongs to the family of heuristics that we consider here. Like many other members of that family, the recognition heuristic for judging city sizes (1) draws on a “natural assessment” of recognition or familiarity, (2) may be endorsed as a deliberate strategy, (3) makes people look smart under some conditions, and (4) will produce systematic errors and biases, because impressions of familiarity and recognition are systematically correlated with factors other than city size, such as number of mentions in the media.