The Stability of Preferences – A Social-Cognition View

Attitudes guide behavior. The socialcognitive approach to decision-making has been building on this assumption for almost a century (Allport, 1935). In this field, the model of reasoned action (Fishbein and Ajzen, 1974) was probably the most influential advance in describing the relation between attitudes and behavior. Accordingly, attitudes are the key predictor of behavioral intentions. They are formally described as a function of a linear integration of evaluations and probabilities (beliefs). The attitudinal part of the model dovetails with the subjectively expected-utility (SEU) approach to risky decision-making (Edwards, 1954). In line with attitude–behavior models, preferences are stable attitudes toward behaviors. After its formation, an attitude can be stored in memory in association with the attitude object (e.g., Fazio, 1990; Wilson et al., 2000; Betsch, 2005). Attitude storage opens the path to stability. If a person re-encounters an attitude object, she can look up her attitude in memory and use it for subsequent judgment and choice. Individual preferences should be stable to the extent that the person relies on stored attitudes. This so-called “file-drawer” notion (Wilson and Hodges, 1992), however, has been challenged by empirical evidence. Attitudinal responses in risky and nonrisky choice domains were shown to be susceptible to a variety of task conditions. Krosnick and Schuman (1988) showed that response order, question wording and format systematically affect attitude judgments irrespective of their subjective importance, extremity, and certainty. In a similar vein, empirical violations of the axioms of decision theory challenged the notion that preferences are stable (e.g., framing effect, Tversky and Kahneman, 1981). The temporary construal approach, in its extreme version, assumes that individuals always construct attitudes on the spot (Schwarz, 2000). In this view, stability is a function of the situation. In stable situations the sample of information (salient stimuli, activated knowledge) is invariant. Consequently, attitudinal judgments and behaviors are expected to be stable. If the situation changes, however, the sample of information changes as well and so will attitudes and behaviors. Accordingly, attitudes are transient constructions. Therefore, preferences cannot be stable. The focus on variability, however, is prone to yield an un-representative picture of human behavior. Numerous examples of behavioral rigidity can be cited, both from everyday observation and controlled studies. Individuals regularly obey norms, repeat their routines, and fancy the same things two days in a row. Correspondingly, there is a bulk of psychological research demonstrating stability, especially in the field of learning. In their famous demonstrations of the Einstellung-Effect, Luchins and Luchins (1959) showed that only a few implementations of a problem-solving strategy suffice to induce a tendency to maintain the strategy even when less costly strategies are adequate. In recurrent risky choices, prior experience fosters maintenance of behavioral options (Betsch et al., 2001) and decision strategies (Bröder and Schiffer, 2006), even following changes in the pay-off structure that render the routine a maladaptive choice. A few behavior repetitions (less than 10) suffice to induce counter-intentional relapse errors in subsequent choice (Betsch et al., 2004). In research on attitude– behavior models, Bentler and Speckart (1979), for instance, suggested including past behavior as an additional predictor in the Fishbein–Ajzen model. How is it possible that there is evidence for both variability and stability in behavior? Many studies on variability employ tasks and judgment domains in which individuals lack behavioral experience. In contrast, studies on stability often assess behavior in recurrent tasks. It is widely acknowledged, however, that experience matters. Decisions based on experience yield different results from those in oneshot situations (e.g., Hertwig et al., 2004). Thus, one might conclude that preferences should be labile in new situations and stabilize with behavioral experience. Note, however, that the behavioral invariance observed in recurrent and experienced-based situations is not a sufficient condition for inferring preference stability. According to attitude theory, preferential stability requires that the attitude remains stable and guides choice. Repetitive experience, however, could merely result in increasing the association between a stimulus and a response. If this were the case, we would expect stability in recurrent situations and variability when situations change in the sense that the learned stimulus is no longer present. Explaining stability in stable situations does not necessitate the assumption of a stable attitude because it is the stimulus situation and not the attitude that might induce stable responses. Stimulus-directed choice is likely to occur if a behavior has been repeated so frequently in the past that it has been “frozen into habit” (James, 1890/1950; Verplanken and Aarts, 1999). Habits are responses that are instantiated automatically upon recognizing the associated stimulus. Indeed, there is evidence that highly frequent behavior repetition can outperform attitudes and intentions (Ouellette and Wood, 1998). On the other hand, the fact that habits are implemented in a stimulus-directed fashion without involving goals and intentions (Wood and Neal, 2007) does not speak against the possibility that attitudes can also become stable with behavior repetition and may systematically guide intentional decisions. Two conditions must be met to justify the assumption that repetition paves the way to preference stability. First, previous behavior must increase the association between an attitude and an object. Accordingly, response latencies for attitude judgments A relapse error occurs if the actor performs a routine behavior against his or her intention to deviate from this routine.