Active learning as a means to distinguish among prominent decision strategies

A long-standing debate in decision making has been whether people rely on very little information for making choices, or weigh and add all available information. We propose a new method to determine whether a non-compensatory (Take-TheBest) or compensatory strategy (Logistic Regression) is more psychologically plausible: by looking at peoples active learning queries. This method goes beyond traditional model selection techniques as it reveals the information people choose to learn early on, which subsequently drives their decisions. We developed active learning algorithms for both Take-The-Best and Logistic Regression, and designed an active learning experiment to distinguish between these models. By letting both models and humans actively learn, we could compare their queries, and found that people follow a rank-based learning strategy in non-compensatory environments, but prefer more certainty-based queries in compensatory environments. We argue that active learning studies provide a promising new methodology to distinguish among decision models.