INFORMS Open Forum

Society of Decision Professionals (SDP) and the Decision Analysis Society (DAS) invites you to join our July 2025 Webinar:

"Risk Matters Less When Options Differ Qualitatively"

Featuring: George Wu, John P. and Lillian A. Gould Professor of Behavioral Science - University of Chicago Booth School of Business

Abstract:

Risk aversion for moderate likelihood gains is perhaps the best-known stylized fact from decision research.  Though studies documenting it have largely focused on decisions concerning money, such behavior is presumed to prevail very generally, across diverse domains.  We investigate the validity of this generalization by contrasting two types of decisions.  In unimodal choices, outcomes differ quantitatively but not qualitatively, rendering inter-option comparisons "apples-to-apples."  Choices between guaranteed and uncertain monetary payoffs are canonical examples; or consider sure receipt of a product versus a chance at several units of the same product.  In crossmodal choices, outcomes differ qualitatively (and sometimes quantitatively), rendering comparisons "apples-to-oranges."  Consider sure receipt of one product versus uncertain receipt of another.  Our studies () reveal a pair of patterns by which, contrary to straightforward generalizations, risk matters less crossmodally.  First, crossmodal preferences exhibit less aversion to fair risk.  Second, crossmodal preferences vary less across risk levels: as risk becomes increasingly unfavorable, they do not develop as much additional distaste for it; and, as it becomes increasingly favorable, they do not develop as much appetite for it.  These patterns of insensitivity jointly engender an interaction: relative to unimodal settings, crossmodal settings yield less aversion to unfavorable and fair risk but more aversion to favorable risk.  To account for this interaction, we present a model in which weighing risk crossmodally requires additional mental operations.  Such extra cognitive requirements introduce error, reduce confidence, and thereby diminish impact. Crossmodal settings thus dilute weighing of risk as a basis for preference construction.