Inferring Fast and Frugal
Heuristics from Human
Ling Rothrock and Alex Kirlik
Much late twentieth-century literature on the judgment and decision making of experienced performers, especially those working in technological settings, suggests that decisions are rarely made according to the prescriptions of highly enumerative strategies such as utility maximization (Savage, 1972) or its multiattribute variant (Edwards & Newman, 1982). Instead, performers seem to rely heavily on one type of heuristic shortcut or another to meet the demands of time stress, uncertainty, information load, and the often abstract manner in which information is displayed in technological contexts. Researchers have proposed various accounts of these shortcuts. The most notable of these, at least in the human factors and cognitive engineering disciplines, has emerged from the naturalistic decision making (NDM) paradigm (Klein, 1999; Klein et al., 1993). From this perspective, the apparent cognitive efficiency of judgment and decision making in these contexts arises from experience providing the performer with a vast storehouse of cases, often allowing a performer to quickly recognize a good solution rather than having to comparatively evaluate a long list of options.
Additionally, heuristic strategies that may initially appear to be vastly oversimplified for meeting the demands of a particular task often yield surprisingly good and robust performance when compared against the performance of more cognitively demanding, enumerative strategies (Connoll, 1988; Gigerenzer & Kurz, 2001; Hogarth, 1981, Kirlik, Miller, & Jagacinski, 1993; Kirlik et al., 1996a, 1996b). A combination of factors is likely to account for this counterintuitive result. Gigerenzer and colleagues (e.g., Gigerenzer & Selten, 2001; Gigerenzer & Goldstein, 1996; Gigerenzer, Todd, & The ABC Research Group, 1999), following Brunswik's lead, have focused on the role of the statistical structure or “causal texture” of the task environment as playing an important role. For example, they have noted that many judgment or decision ecologies have cue-criterion structures that allow few or perhaps even one cue to provide the basis for sound judgment (see chapter 18 for a concrete example). For instance, knowing that a particular U.S. city has a major league baseball team and another one does not is a highly reliable predictor of relative city population. In addition, one might think that by taking more cues into account one would successively improve the quality of a prediction such as this, but as the work of Gigerenzer and colleagues has demonstrated, this is not always the case. As in linear regression, increasing the number of cues or predictor coefficients