Framing effects: An analytic–holistic perspective

Under what conditions, why, and for whom are framing effects most likely? In this paper, we build on the existing literature (e.g., Chaiken, 1987; Epstein, Lipson, Holstein, & Huh, 1992; Evans & Over, 1996; Fiske & Neuberg, 1990; Payne, Bettman, & Johnson, 1988; Simon, 1956; Sloman, 1996; Stanovich & West, 2000), in providing answers to these questions. We hypothesized that individuals who engage a decision task with an analytic/systematic versus holistic/heuristic processing style are especially insensitive to the influence of framing effects. Therefore, we predicted that the way in which a decision is framed should have a relatively weak influence on those who were either induced (Experiment 1) or predisposed (Experiment 2) to adopt a predominantly analytic/systematic versus holistic/heuristic processing style. The results of both experiments supported this position. Over the last few decades, there has been considerable interest in examining how the framing of a decision task influences the types of decisions that people make. Framing effects refer to ‘‘the finding that decision makers respond differently to different but objectively equivalent descriptions of the same problem’’ (Kuhberger, 1998, p. 150). Research has examined both the theoretical components of framing (e.g., Fagley & Miller, 1990; Tversky & Kahneman, 1981) as well as its application for a variety of different areas including medical decisions (e.g., O’Connor, Pennie, & Dales, 1996), monetary decisions (e.g., Fagley & Miller, 1997), and taxes (e.g., Highhouse & Paese, 1996). Predictions concerning framing are typically derived from prospect theory. According to prospect theory (Kahneman & Tversky, 1979, 1984; Tversky & Kahneman, 1981), the decision making process is dichotomous, consisting of two phases. The initial phase involves editing of the available options. This initial phase is characterized by the simplification and reorganization of the available options. The options are first coded as gains or losses, relative to a neutral reference point. They are then broken down into a simpler form and given an individuating code to produce streamlined versions of the original options. The subsequent phase involves evaluation of the edited options. In this phase, the subjective values and the weighted probabilities for each of the options are evaluated in light of the prior editing phase. The subjective values and probabilities are then integrated into a single value and contrasted relative to each other. Framing effects in the evaluation phase are depicted by the S-shaped value function. The shape of this function is concave for gains and convex for losses. This depicts prospect theory’s prediction of risk-aversion for gains and risk-seeking for losses. That is, an individual in the concave curve should prefer 100 dollars for certain rather than a 50/50 chance to win 200 or win nothing. An individual in the convex curve should prefer a 50/50 chance to lose 200 dollars rather than losing 100 dollars for certain. Although there are several different types of framing manipulations (see Levin, Schneider, & Gaeth, 1998); the most widely tested example involves the Asian disease problem (e.g., Tversky & Kahneman, 1981). In the Asian disease problem the same basic message is used in both the gain and loss frames. Participants are told that the US is preparing for the outbreak of an unusual Asian disease, which is expected to kill 600 people. They are then presented with two options to combat the disease, and asked to choose between them. When framed in terms of gains, the two options consist of: 200 people will be saved for certain (risk-free option) contrasted with a 1/3 chance that 600 people will be saved and a 2/3 chance that no one will be saved (risk-seeking option). When framed in terms of losses, the two options consist of: 400 people will die for certain (risk-free option) contrasted with a 1/3 chance that no one will die and a 2/ 3 chance that 600 people will die (risk-seeking option). As a variety of different studies have shown (see Kuhberger, 1998 for review), participants tend to choose the risk-safe option (200 saved for certain) when the problem is framed in terms of gains and the risky option (1/3 chance that no one will die and a 2/3 chance that 600 will die) when the problem is framed in terms of losses. The propensity for decision-makers to choose risk-averse options, when Asian disease type problems are framed in terms of gains, and risk-seeking options when framed in terms of losses, is inconsistent with the rational predictions of expected utility theory (e.g., Von Neumann & Morgenstern, 1944). According to this view, the way in which the decision is framed should not change the expected utility of either the risk-seeking or risk-averse options. If, because of a person’s utility function, the risk-averse option is chosen in the gains condition, this option should also be chosen in the losses condition; conversely, if the risk-seeking option is chosen in the gains condition, it should also be chosen in the losses condition. Recent reviews of framing (e.g., Kuhberger, 1998; Levin et al., 1998) have concluded that although there is a moderately strong framing effect for manipulations that follow the Asian disease paradigm, framing effects are not always obtained (e.g., Bless, Betsch, & Franzen, 1998; Highhouse & Paese, 1996; O’Connor et al., 1996). Therefore, a key question is under what conditions, and for whom are framing effects most likely (Fagley & Miller, 1990; Kuhberger, 1998; Levin et al., 1998; Stanovich & West, 2000). Framing and a dual-route analysis Recently, Stanovich and West (2000) have used the general assumptions of a family of dual-process theories of reasoning (e.g., Epstein et al., 1992; Evans & Over, 1996; Sloman, 1996) to explain why some individuals behave in a maximizing/rational way whereas others do not. They propose that, although the exact properties of these dual-process accounts are not always identical, there are several conceptual similarities that constitute a common class of assumptions. These views all assume that there are two different processing routes and that each route leads to different types of construals. One route (system 1 in Stanovich and West’s terminology) is relatively automatic and holistic and leads to an automatic contextualization of problems whereas the other route (system 2) involves a more controlled and analytic processing style and serves to decontextualize and depersonalize problems. Rather than just engaging in a discreet analysis of the information, individuals who use a holistic processing style rely on contextual cues that allow them to rely upon internal representations of the problem. This, in turn, allows them to make inferences about the issue or task without detailed scrutiny of the material. This type of processing should be especially sensitive to contextual cues, such as how the problem is framed. Stanovich and West (2000) argue that individuals with a high level of analytic intelligence are more likely to engage in system 2 processing, and that, at times, individuals with especially high levels of cognitive ability will have the requisite computational ability to override the influences of system 1. These individuals are able to abstract difficult problems into canonical representations that are devoid of context, thereby freeing them from mistakes that result from the erroneous use of nonrelevant information, such as the way the problem is framed. Individuals engaging a task with this type of analysis initially perform a discreet analysis on each of the options within the task. Accordingly, the analysis of each alternative should then be aggregated and the option with the greatest numeric potentiation should be systematically chosen. This type of analysis, when applied to a decision task containing options of equal values, should yield similar results, independent of the framing of the task. Stanovich and West (2000) use their model to explain why several authors (e.g., Johnson-Laird, 1999; JohnsonLaird & Byrne, 1993) propose that the way in which perceivers search for a model is not governed by an ordered system of principles. Building on their system 1/ system 2 distinction, Stanovich and West (2000) conclude that some of the variability in perceivers’ search activities can be accounted for by considering individual variations in cognitive ability and task construals (see also Stanovich & West, 1999). They suggest that the search process appears to not be determined by an organized system because individuals differ, for example, in their propensity to use different cognitive processing styles, such as system 1 and system 2 styles of thought, as well as in their tendencies toward cognitive closure/structure (e.g., Kruglanski, 1989; Neuberg & Newsom, 1993). Therefore, each individual’s style of epistemic regulation may be relatively consistent, despite the existence of inter-individual variability in cognitive styles. In their analysis, Stanovich and West (2000) concentrate on individual differences in intellectual/cognitive abilities and not on how contextual and cognitive style differences impact the likelihood and strength of framing effects. Their distinction, however, between system 1 and 2 processing is consistent with several social psychological approaches (e.g., Chaiken, 1987; Epstein et al., 1992; Fiske & Neuberg, 1990; Petty & Cacioppo, 1986) that have been used to examine how contextual and cognitive style differences influence an individual’s processing style and ultimately social judgment. Consequently, these models are especially relevant to the question examined in the present paper; namely how are framing effects influenced by contextual and cognitive style differences. To answer this question, we use the central assumptions of Chaiken’s (1987) systematic/heuristic dual-process model as well as previous theorizing that implicates cognitive effort in the decision process (e.g., Payne et al., 1988; Simon, 1956). Because holistic/heuristic processing is less effortful than analytic/systematic processing, cognitive, and motivational factors influence the use of these two processing styles (e.g., Chaiken, 1987). Holistic/heuristic processing normally occurs when an individual has low levels of motivation or ability whereas, because analytic/systematic processing is significantly more effortful than holistic/heuristic processing, it only occurs when individuals are both willing (having sufficient motivation) and able (having sufficient capacity and capability) to perform the task at hand. Therefore, in situations where cognitive ability is not constrained, motivational factors, such as the personal relevance of the decision task, emerge as the determining cause for effort allocation. Specifically, as the relevance of the decision increases, so too does the amount of effort that an individual is willing to expend on the decision task; and as the amount of task-related effort increases, so too does the likelihood of analytic/systematic processing. Overview of Experiments 1 and 2 In Experiment 1, we manipulated the self-relevance of the decision context to alter individuals’ willingness to expend resources (cognitive effort) on the decision task. We expected individuals to adopt a predominately analytical style of thinking when they worked on a personally relevant decision task whereas we expected them to adopt a predominantly holistic/heuristic style when the decision task was personally irrelevant. Consequently, framing effects should be especially weak when participants work on a self-relevant task but not when they work on a personally irrelevant one. In Experiment 2, we took advantage of the fact that there are individual differences in the use of analytical/systematic versus holistic/heuristic processing (e.g., Zenhausern, 1978) by comparing the decisions of individuals who were predisposed to use either an analytic or holistic style of thinking. Converging evidence for a holistic/analytical analysis of framing will be obtained, if, across both experiments, we observe weaker gain/loss framing effects when individuals engaged in an analytic as opposed to a holistic style of thinking.