Automated Choice Heuristics

In his Nobel Prize address, Herbert Simon urged choice theorists to stop pretending that actual choices can be predicted from theoretical models of optimal choice. He argued that any descriptively adequate account of human decision making must make contact with the actual psychological processes that are involved and that "the neoclassical ambition of avoiding [this] necessity is unrealizable" (1978, p. 507).

Many choice theorists have taken up Simon's challenge. They have sought to identify the various "heuristics" people use to simplify choice – the procedures they use to limit the amount of information that is processed or the complexity of the ways it is combined. In a prototypical study, a respondent might be asked to make several hypothetical choices among apartments, whose attributes (e.g. monthly rent, miles from work, square footage) are listed as rows of numeric values (e.g. $560, 12, 900, respectively). The respondent's choice processes are then either inferred from the choices she makes or are more directly observed through "process tracing" methods that use computerized displays to track the attributes and options that the respondent considers, the order in which they are considered, and, sometimes, the time spent pondering each piece of information.

On the basis of process tracing methods, verbal protocols, introspection, and theory, many different choice heuristics have been postulated (for reviews, see Payne, Bettman, & Johnson, 1993; Gigerenzer, Todd, & the ABC Group, 1999; Gigerenzer, Czerlinski, & Martignon, Chapter 31, this volume; Gigerenzer & Goldstein, 1996). These choice heuristics, such as "elimination by aspects" (Tversky, 1972), are conscious strategies that are intentionally designed to simplify choice. When a decision maker is presented with a matrix of numbers summarizing the attributes of six different apartments, nothing spontaneously "happens"; no intuitive computation generates an impression of which option is best. The decision maker must consciously decide how to process that information. These deliberate choice heuristics are viewed as heuristics by the people who use them – they would describe what they are doing as a simplifying strategy. Indeed, the "adaptive decision maker" envisioned by some choice theorists (Payne, Bettman, & Johnson, 1993) is not only aware that he is using heuristics, but modifies them strategically as choice conditions change. The deficiencies of these heuristics are not lost on decision makers – they often recognize when they are ignoring potentially relevant information, but may regard their strategies as appropriate given their computational limitations or their desire to conserve effort.

These deliberate choice heuristics differ substantially from the judgmental heuristics of the "heuristics and biases" research program, which are largely based on impressions that occur automatically and independently of any explicit judgmental goal (Tversky & Kahneman, 1983, Chapter 1, this volume). For example, when presented with a description of a political activist and the category bank teller, people notice the dissimilarity spontaneously. This rapid and intuitive judgment may, in turn, be used to conclude that the person is unlikely to be a bank teller by a process that Kahneman & Frederick (Chapter 2 this volume) term attribute substitution. Importantly, however, the people making the judgment do not typically view this process as a heuristic. They are not deciding to use resemblance in favor of some other procedure for judging probability. They do not see it as an effort conserving procedure, or a simpler version of another procedure they might instead perform. Using resemblance to judge probability is called a heuristic (the "representativeness heuristic") only by reference to an "outside" normative standard of judgment, which requires that these probability judgments reflect considerations other than resemblance alone.

The distinctions between the "deliberate" choice heuristics (e.g. elimination by aspects) and the "automatic" judgmental heuristics (e.g. representativeness) largely reflect the different types of stimuli used in choice tasks and judgment tasks. Research on choice heuristics was conducted with an eye toward the computational and memory limitations that Simon repeatedly emphasized, and focused on how people cope with complexity; how they make reasonable decisions when the amount of information confronting them exceeds the amount they can comfortably process. The experimental stimuli used by these researchers are often abstract and do not evoke any type of intuitive impression. Consequently, respondents must rely on some type of analytic solution (though the "analysis" may be rather crude). In contrast, much of the research on judgment heuristics was guided by the analogy with perception, and presented respondents with more concrete stimuli (e.g. personality descriptions) that evoke an immediate impression upon which the requested judgment can be based. In the shorthand used by Kahneman and Frederick (Chapter 2, this volume), traditional judgment heuristics are System 1 heuristics – they result from cognitive processes that are rapid and not entirely controllable – whereas traditional choice heuristics are System 2 heuristics – they result from slower and more deliberate mental processes.

Although the prototypical choice heuristic is considerably more deliberate than the prototypical judgmental heuristic, there are situations in which choices are governed by (and simplified by) intuitive impressions. This chapter discusses two such automated choice heuristics: "choosing by liking" – basing choice on an immediate affective evaluation; and "choosing by default" – choosing the option that first comes to mind. These two automated choice heuristics are akin to the judgment heuristics of the heuristics and biases literature – they are governed by rapid and intuitive processes, they are relatively immune to introspection, and their associated biases may not be recognized by the people who use them.