This paper studies stochastic choice in social contexts, and explores the difference between inequity-averse and image-conscious preferences, in particular, shame aversion. This paper studies a class of additive perturbed utility (APU) introduced in Fudenberg et al. (2015). We find that the general class of inequity-averse preferences correspond to the case of item-invariant APU, and that image-conscious preferences correspond to the case of menu-invariant APU. Under some conditions, the difference between inequity aversion and shame aversion is discussed through the violation of Regularity, one of the most well-known properties of stochastic choice.
The full paper is available at TCER (link). Please see the latest version in SSRN.
Attribute-Based Inferences and Random Limited Consideration: A Representation Result
This paper presents an axiomatic model of random limited considerations under attribute-based inferences. To characterize the model, this paper takes the framework of preferences over menus. The key axiom characterizes a model of random consideration sets, which is an extended and generalized version of the reference-dependent choice model of Ok et al. (2015) in the sense that (i) consideration sets are formed under attribute-based inferences randomly, and that (ii) reference points themselves can be chosen from menus. The two extensions make it possible to allow for not only the Attraction Effect, but also the Compromise Effect.
This paper investigates preferences over menus, i.e., sets of alternatives, to analyze a decision maker who randomizes alternatives in a menu. We identify preferences for randomization, by studying anticipated utility from the viewpoint of subjective uncertainty. The effect of randomization in one's mind stems from several cognitive or psychological effects ranging from complements between alternatives to aversion to timing of earlier decision-making. The main axiom in our analysis is a monotonic condition for preferences for randomization, named as Randomization. We show that Randomization, along with other standard axioms, axiomatically characterizes the main result, that is, a random anticipated utility representation. In the model, the decision maker's subjective belief for the effect of randomization is uniquely identified. In addition, we characterize preferences for a desire to randomization, and an aversion to randomization, respectively.
The axioms explained in this proceeding is not sufficient for the utility representation. I fix the results, and present the formal model of deliberate randomization.