This research was supported by grants from the Israeli Science Foundation (327/10) and the Chicago Wisdom Project (John Templeton Foundation, Subaward FP050019-B). We would like to thank Robin Hogarth and Ed Vul, who both had an important influence on this work. We are also grateful to Shlomi Sher, Mike Yeomans and two anonymous reviewers for their comments, and to Naomi Goldblum for her editorial help.
Previous research has shown that a speaker's choice between logically equivalent frames is influenced by reference point information, and that listeners draw accurate inferences based on the frame. Less clear, however, is whether these inferences play a causal role in generating attribute framing effects. Two experiments are reported, which suggest that frame-dependent inferences are sufficient to generate attribute framing effects, and that blocking such inferences may block framing effects. Experiment 1 decomposed the typical framing design into two parts: One group of participants saw a target described in one of two attribute frames and reported their estimates (inferences) of the typical attribute value. These estimates were then given to a second group of yoked participants, who evaluated the target. Although this latter group was not exposed to different attribute frames, they nevertheless exhibited a "framing effect" as a result of receiving systematically different inferences.In contrast, Experiment 2 shows that experts-who are familiar with an attribute's distribution and are therefore less likely to draw strong frame-based inferences-exhibit a diminished framing effect. Together, these findings underscore the role of inferences in the generation and attenuation of attribute framing effects.McKenzie, 2014) and intransitive choice behavior (Müller-Trede et al., 2015), point to an important role for inferences in context and framing effects.
There is evidence that betting on longshots increases in the last race of a day of horse racing. Previous accounts have assumed that the phenomenon is driven by bettors who have lost money and are trying to recoup their losses. To test this assumption of "reference dependence," three laboratory experiments simulated a day at the races: In each of several rounds, participants chose either (i) a gamble with a small probability of a large gain and a large probability of a small loss (the "longshot") or (ii) a gamble with a moderate chance of a small gain or a small loss (the "favorite"). The first two experiments employed a game played for points, while a third experiment included monetary incentives and stimuli drawn from a real day of racing. These experiments provide a clear demonstration of the last race effect in a laboratory setting. However, the results indicate that the effect is largely reference independent: Participants were more likely to choose the longshot in the last round regardless of whether, and how much, they had won or lost in previous rounds. Winning or losing, bettors prefer to "go out with a bang" at the end of a series of gambles.1 This phenomenon is not limited to horse races using the parimutuel betting system. In the UK, bookmaker-based odds are the norm, and a favoritelongshot bias has also been shown there (Dowie, 1976).
Preferences must be constructed at least some of the time. This, by itself, is not problematic for rationality. At issue is whether the construction is done in a reasonable manner. The common view is that preference construction violates coherence principles that are basic requirements of rational choice. However, traditional coherence principles are static and implicitly assume that the choice context provides no relevant information. In lab experiments, decision makers often evaluate or choose between options that are unfamiliar or even fictitious, and they may look to the context for choice-relevant cues that help them update their beliefs and construct their preferences. We review evidence that a number of apparent "biases" in decision making stem from adaptive sensitivity to subtle contextual cues. These context effects are dynamically coherent, in that preference-updating is coordinated with reasonable context-dependent belief-updating. This perspective on preference construction not only provides a different view of the psychology and rationality of decision making, it also suggests a different approach to choice architecture. Whereas the traditional nudge approach tries to engineer specific decision outcomes, often by rerouting apparent biases so that they point in desirable directions, the present approach seeks to facilitate processes in order to help people make rational decisions.
Several apparent violations of transitivity have been reported in the literature on decision making. However, these effects have been shown to be compatible with random preference models, in which preferences are transitive at each point in time but vary at random over time. Such models imply that choice proportions will conform to a set of conditions called the triangle inequalities, and no clear triangle inequality violations have been empirically demonstrated to date. This article examines a broader class of choice models—“context-sensitive preference models”—in which the current and prior history of choice contexts can systematically influence decision makers’ stochastic preferences. These models generate violations of the triangle inequalities even when preferences are always transitive. Furthermore, the article develops an analysis of decision making under incomplete information, in which rational decision makers draw inferences from the present choice context, but have limited memory for past contexts. It is shown that such decision makers can exhibit intransitive choice cycles of arbitrary magnitude as a result of context-dependent switching between transitive preference orders. Two experiments test the model’s predictions, and clear violations of the triangle inequalities are observed.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.