Cardinal Utility and Contrast in Models of Choice and Response Time

Abstract: Response times can help reveal preferences when choices are noisy. This paper introduces a novel framework to analyze models that predict choices and response times from the utility of the alternatives. Decomposing models based on which choice and response time distributions they generate and how these distributions are normatively interpreted, I uncover a key elasticity - the contrast level - that drives their behavioral predictions. However, the contrast level is arbitrarily set and not identifiable by current calibration methods. I propose a new calibration method that I illustrate on practical examples with data simulated from a Drift Diffusion Model. My results shed light on the normative implications of a wide class of models from cognitive science and offer guidance on their use in economics. They are relevant wherever random utility models are currently used, for instance demand models or discrete choice experiments.

Download Paper (PhD thesis version)

Presented at: Lancaster 2025 Psychoeconomics Workshop, ESEM 2024, MathPsych 2024, DICE Behavioural Coffee Seminar, Paris Junior Experimentalist Meeting (design)