Dorison, C. A., & Heller, B. H. (2022).
Journal of Experimental Psychology:
General. Advance online publication.
Abstract
A large interdisciplinary body of research on human judgment and decision making documents systematic deviations between prescriptive decision models (i.e., how individuals should behave) and descriptive decision models (i.e., how individuals actually behave). One canonical example is the loss–gain framing effect on risk preferences: the robust tendency for risk preferences to shift depending on whether outcomes are described as losses or gains. Traditionally, researchers argue that decision makers should always be immune to loss–gain framing effects. We present three preregistered experiments (N = 1,954) that qualify this prescription. We predict and find that while third-party observers penalize decision makers who make risk-averse (vs. risk-seeking) choices when choice outcomes are framed as losses, this result reverses when outcomes are framed as gains. This reversal holds across five social perceptions, three decision contexts, two sample populations of United States adults, and with financial stakes. This pattern is driven by the fact that observers themselves fall victim to framing effects and socially derogate (and financially punish) decision makers who disagree. Given that individuals often care deeply about their reputation, our results challenge the long-standing prescription that they should always be immune to framing effects. The results extend understanding not only for decision making under risk, but also for a range of behavioral tendencies long considered irrational biases. Such understanding may ultimately reveal not only why such biases are so persistent but also novel interventions: our results suggest a necessary focus on social and organizational norms.
From the General Discussion
But what makes an optimal belief or choice? Here, we argue that an expanded focus on the goals decision makers themselves hold (i.e., reputation management) questions whether such deviations from rational-agent models should always be considered suboptimal. We test this broader theorizing in the context of loss-gain framing effects on risk preferences not because we think the psychological dynamics at play are
unique to this context, but rather because such framing effects have been uniquely influential for both academic discourse and applied interventions in policy and organizations. In fact, the results hold preliminary implications not only for decision making under risk, but also for extending understanding of a range of other behavioral tendencies long considered irrational biases in the research literature on judgment and decision making (e.g., sunk cost bias; see Dorison, Umphres, & Lerner, 2021).
An important clarification of our claims merits note. We are not claiming that it is always rational to be biased just because others are. For example, it would be quite odd to claim that someone is rational for believing that eating sand provides enough nutrients to survive, simply because others may like them for holding this belief or because others in their immediate social circle hold this belief. In this admittedly bizarre case, it would still be clearly irrational to attempt to subsist on sand, even if there are reputational advantages to doing so—that is, the costs substantially outweigh the reputational benefits. In fact, the vast majority of framing effect studies in the lab do not have an explicit reputational/strategic component at all.