Abstract
In the last years, research on risky choice has moved beyond analyzing choices only. Models have been suggested that aim to describe the underlying cognitive processes and some studies have tested process predictions of these models. Prominent approaches are evidence accumulation models such as decision field theory (DFT), simple serial heuristic models such as the adaptive toolbox, and connectionist approaches such as the parallel constraint satisfaction (PCS) model. In two studies involving measures of attention and pupil dilation, we investigate hypotheses derived from these models in choices between two gambles with two outcomes each. We show that attention to an outcome of a gamble increases with its probability and its value and that attention shifts toward the subsequently favored gamble after about two thirds of the decision process, indicating a gaze-cascade effect. Information search occurs mostly within-gambles, and the direction of search does not change over the course of decision making. Pupil dilation, which reflects both cognitive effort and arousal, increases during the decision process and increases with mean expected value. Overall, the results support aspects of automatic integration models for risky choice such as DFT and PCS, but in their current specification none of them can account for the full pattern of results.
Highlights
In many every day decisions individuals choose between options with different outcomes, each of which realizes with a certain probability
Given our results and evidence from previous studies, simple non-compensatory models such as the PH, LEX, minimax, or maximax heuristic do not seem to be appropriate to predict search behavior and processes involved in risky decisions in general
The same holds for the serial implementations of expected utility (EU) models in the form of WADD
Summary
In many every day decisions individuals choose between options with different outcomes, each of which realizes with a certain probability. Research in risky choice has a long tradition, and several models have been developed to account for the wealth of identified choice anomalies and data observed on risky choices in the lab and the field. Many of these models are extensions and refinements of the expected utility (EU) Model (von Neumann and Morgenstern, 1944; Edwards, 1954; Savage, 1954) assuming that some kind of integration of payoff and probability of outcomes drives individuals’ choice. One important limitation of these models, is that they predict choices only and remain largely silent concerning the processes underlying choice (Luce and Raiffa, 1957; Luce, 2000)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have