Title: Psychologically interpretable differences in decision making under uncertainty
Abstract:
Existing models of decision making under uncertainty typically fit a limited number of parameters to explain behavior in choice tasks. In a notable deviation, two Choice Prediction Competitions found that a model based on many behavioral tendencies (e.g., sensitivity to the signs of prospects) produced superior predictive accuracy. In two datasets (an existing dataset of 27,630 choices and a new dataset of 40,100 choices), I compare the predictive performance of fixed effects models and models adding random effects to these behavioral features, before focusing on psychologically interpreting these relationships. Specifically, I use hierarchical linear models to show that i) there are individual differences in sensitivity to such behavioral features that characterize aspects of decision making processes, ii) these are stable across time, sample population, problem, and context (financial vs. abstract), and iii) that other individual differences (e.g., cognitive reflection) predict these sensitivities. I conclude by discussing the implications of these results for deploying models of decision making under uncertainty that are both predictive and explainable in important decision contexts.
Zoom Link: https://us02web.zoom.us/j/82097577813