Choice-induced preferences in the absence of choice: Evidence from a blind two choice paradigm with young children and capuchin monkeys

Abstract

In two studies, we provide a test of the revealed-preferences account of choice-induced preferences. To do so, we introduce a blind two-choice task in which preferences cannot guide choices. Children chose between two similar objects while ignorant of the objects’ identities, and then chose between the rejected alternative and a third similar object. Monkeys were given an illusion of choice between two similar objects, and then chose between the rejected object and a third similar object. Both children and monkeys preferred the third object, indicating that they devalued the rejected object. This response pattern did not occur when the children and monkeys were not given the opportunity to choose between the two initial items. These results provide evidence against a revealed-preferences account of choice-induced preferences and demonstrate that the process of making a choice itself induces preferences.

Introduction

Common sense suggests that people make choices based on their preferences, but a long literature in social psychology suggests that choices might actually induce preferences. After choosing between equally-attractive alternatives, an individual’s liking for the chosen option apparently increases (e.g., Brehm, 1956, Jarcho et al., submitted for publication, Lieberman et al., 2001, Lyubomirsky and Ross, 1999, Steele, 1988) and liking for the rejected option apparently decreases (e.g., Brehm, 1956, Egan et al., 2007, Jarcho et al., submitted for publication, Lieberman et al., 2001, Lyubomirsky and Ross, 1999, Steele, 1988).

Over the past few decades, the phenomenon of choice-induced preferences has generated much discussion in the field of social psychology, but the debates surrounding this phenomenon have generally taken two rather different forms. The first form of debate over choice-induced preferences has surrounded the nature of the mechanisms that underlie this phenomenon. The original choice-induced preference paradigm was developed as a method for studying the phenomenon of cognitive dissonance reduction (e.g., Brehm, 1956). Since this initial work, a number of other mechanisms have been proposed to account for choice-based preference changes, some of which are consistent with a motivation to reduce cognitive dissonance. However, one of the cleaner ways to distinguish among the explanations is by whether they claim a central role of the self.

Generally, researchers have proposed two classes of candidate mechanistic explanations: those that posit a central role for the self, and those that do not. The first class of explanations, which includes models like self-consistency theory (e.g., Aronson, 1968) and self-affirmation theory (e.g., Steele, 1988), argues that choice-induced preference changes result from discrepancies between one’s self concept and one’s decisions. Under this class of explanations (which we refer to here as “self-based models”), bad decisions somehow pose a threat to one’s self-concept; people therefore tend to change preferences so that their decisions appear to reflect more a consistent or positive self. The second class of explanations—which we’ll refer to as “non-self-based models”—tend to involve processes that don’t require sophisticated self-concepts. Instead these models explain choice-induced preferences as the result of other more general kinds of processing. Such models have argued, for example, that choice-induced preferences result from network-level attempts to reconcile preference inputs and behavioral outputs (e.g., Shultz and Lepper, 1996, Van Overwalle and Jordens, 2002) or general processes involved in inferring people’s preferences and behaviors (e.g., Bem, 1967, Bem, 1972). Critically, this second class of models does not require that decision-makers have either a rich self-concept or advanced meta-representational capacities.

In order to distinguish between these two classes of models, researchers have recently begun investigating whether two populations whose self-concepts appear to be less elaborated than those of human adults—non-human primates and young children (e.g., Egan et al., 2007)¹—exhibit choice-induced preferences. If self-concepts are required for choice-induced preferences, as self-based models maintain, then non-human primates and children should not demonstrate such preferences. In contrast to this prediction, Egan and colleagues (2007) observed that children and non-human primates devalue alternatives that they have previously rejected, changing their preferences to fit with their choices. These results therefore provide support for the second class of models, ones that explain choice-induced preferences without requiring rich self-concepts. In this way, comparative-developmental evidence has provided a novel and useful window into the debate regarding the mechanisms underlying choice-induced preferences.

The present paper attempts to use similar comparative-developmental evidence as a window into a second (and more recent) kind of debate that has surrounded choice-induced preferences: namely, whether this behavioral phenomenon actually exists in the first place. In the past two years, Chen and colleagues (Chen, 2008, Chen and Risen, 2009) have argued that choice-induced preferences are actually methodological artifacts of the experimental procedures typically used in preference studies. They argue that rather than inducing preferences, the act of making a choice merely reveals preexisting preferences towards choice alternatives. To illustrate how this revealed-preference account works, consider the specific methods used by Egan et al. (2007). In this study, children and monkeys were presented with three items (A, B, and C) that were assumed to be equally-preferred based on prior testing. Participants were first offered a choice between A and B, and then a subsequent choice between the rejected alternative (for instance, B) and C. Participants tended to choose C, consistent with the view that the act of rejecting B lowered its value. But suppose now that participants actually had a ranked preference ordering for the three items, one that was not revealed through prior pre-testing. Under this scenario, an individual would choose A over B because of a prior preference for A over B. If so, then her preference ranking must be either [A, B, C], [A, C, B], or [C, A, B]. With such preexisting preferences, there is in fact a 2/3 probability that the individual will select C over B in a subsequent test condition. In this way, Chen and colleagues’ revealed-preference account can explain the results of Egan et al. (2007) without positing any effect of choice on preference. Chen, 2008, Chen and Risen, 2009 have argued that the results from a range of studies on choice-induced preferences fall prey to the same alternative explanation; such results, it is argued, reflect an individual’s true preferences between the available alternatives rather than derogation of unchosen alternatives (but see Sagarin & Skowronski, 2009, for a critical discussion).

In the current paper, we attempt to use comparative-developmental evidence to empirically refute this revealed-preferences account. To do so, we develop a “blind” two choice paradigm, modeled on the two choice paradigm used with children and monkeys described above (Egan et al., 2007). The key feature of the blind two choice paradigm presented here is that an individual’s initial choice is designed such that preferences cannot play a role in choice behavior. We predict that if the mere act of making a choice creates preferences, then participants should choose the third object over the one they rejected, albeit blindly. This effect should be present relative to a parallel No Choice condition in which individuals are given either A or B, but do not choose between the two.

In the first study, we investigate children’s preferences for a blindly rejected toy compared with a third toy. In the second study, we investigate capuchin monkeys’ preferences for blindly rejected food items. Note that both of these two populations are useful for this sort of “blind” test because they lack the demand characteristics that might make similar procedures problematic in adult human populations.