Beyond Lotteries: An Affect-Based Computational Framework For Modeling Risky Choices with Nonmonetary Outcomes

The development of formal models of decision making under risk has been shaped largelyby decisions between options with monetary outcomes. The most prominentmodel—cumulative prospect theory (CPT)—is good at describing choices betweenmonetary lotteries, but performs less well with nonmonetary and nonnumerical outcomes(e.g., medications with possible side effects). We suggest that affective processes, which arenot considered in CPT, play a larger role in nonmonetary than in monetary choices, andpropose two psychologically motivated modifications to CPT’s modeling framework tocapture these differences: (a) using affect ratings rather than monetary equivalents torepresent the subjective value of nonmonetary outcomes (affective valuation); and (b)allowing the probability weighting of an outcome to depend on the amount of affecttriggered in a choice problem (affective probability weighting). We compared model variantsof CPT implementing the proposed modifications in four empirical datasets (totalN = 240). For choices between options with negative nonmonetary outcomes (medicationswith possible side effects), these modifications substantially improved model performancerelative to the standard implementation of CPT. The same did not hold for monetarychoices. Further, an eye-tracking study on nonmonetary choice (N = 68) provided evidencefor two key behavioral and cognitive predictions of affective probability weighting—namely,that risk aversion increases and attention to probability information decreases as theaffective value of the worst outcome in a choice problem increases. Our work integratesprevious ideas on how affect guides and modulates preference construction within acomputational model and delineates an important context in which these mechanismsapply.