Using eye-tracking techniques to understand the role of attention on choice and reversal

A preference reversal (PR) refers to behavior that violates revealed preference or is simply incoherent – i.e., not explainable by a rational ordering. In a classical PR experiment, for instance, participants often exhibit greater risk aversion in a Choice-based revelation procedure than in an Evaluation-based one, i.e., choose the safer of two gambles but express a higher monetary valuation for the riskier. We conjecture that PRs are partly due to the interaction between attention and task mode, and explore three compatible explanations using eye-tracking techniques. Explanation 1 says that difficult tasks require more time to be performed without ‘mistakes’. Those who pay scarcely more attention to Evaluation than Choice, therefore, are more likely to act incoherently. Our data corroborate such a prediction. Explanation 2 assumes that the subjective value of (i) a bet or (ii) any of its attributes (prize and winning probability) depends on the attention paid to it. PRs occur when people allocate attention to different elements across tasks. In line with recent models of drift-diffusion, we find evidence consistent with point (i): a higher focus on the safer bet during Choice predicts PRs. In contrast, little evidence supports the idea (ii) that the share of fixations on probabilities versus prizes influences behavior or PRs. Explanation 3, finally, states that the nature of the tasks may affect the comparisons people make between the options, which are relevant for behavior. For instance, the cognitive difficulty of pricing a bet in Evaluation could distract attention from the relative risk across bets, thus reducing risk aversion. In our design, both bets are visible on the computer screen in both tasks, and subjects make substantially more transitions between bets in Choice. Yet this is observed among all participants, not only the reversers.We are grateful to […], and participants at seminars in the IPP-CSIC […] for helpful comments and suggestions. We also gratefully acknowledge financial support from the Spanish Ministry of Economics, Industry and Competitiveness through the research project ECO2017-82449-PN