How Risk Attitudes Shape Consumer Preferences: An Interpretable Learning Framework

Consumers’ risk attitudes shape how they evaluate and trade off product attributes, yet most preference estimation approaches overlook risk attitudes, assuming risk neutrality. The resulting preference structures may fail to capture the heterogeneity essential for effective market segmentation and product strategy. We propose a risk-driven preference learning framework that reconceptualizes risk attitude as an integral factor shaping preference structures. Its core is a risk-contingent value function in which risk attitude determines how attributes are traded off. This function infers risk attitudes through deviations from full compensation among attributes under risk neutrality, endogenizing risk-attitude inference within preference estimation, and yields preference structures from behavioral data. The method also identifies distinct preference subtypes within the same risk-attitude category. Validated on 114,317 vehicle reviews from Edmunds, an automotive e-commerce platform, the approach achieves strong in-sample fit to behavioral data. Results show that risk-averse, risk-neutral, and risk-seeking consumers exhibit different preference structures, and the risk-averse group shows pronounced internal heterogeneity. The proposed framework offers a structurally interpretable tool for intelligent decision support in consumer segmentation and product strategies.