This week, we had Eric Shulz visiting from UCL and giving the following talk:

Active learning as a means to distinguish among prominent decision strategies

There has been a long-standing debate about whether people rely on single cues for decision making or use more integrative strategies that weigh and add information. A key factor for determining which decision strategy is adapted is the degree to which a learning environment is compensatory. In the current study, we propose a new way to tell apart one-reason decision making (Take-The-Best) and more integrative decision mechanisms (Regression) by the means of active learning. We argue that if a cognitive agent has learned to obey a specific decision making procedure in a pre-defined environment, then the way she selects information over time should reflect that very same architecture. Based on an entropy minimizing active learning algorithm we set out to test this assumption. In an active learning experiment we introduce environments of varying “compensatoriness” by a stick breaking process and let both models and humans actively learn. We then compare queries of a rank and weight-based learning algorithm with participants active queries. Results show that people seem to follow a rank-based learning strategy in non-compensatory environments, and -surprisingly – prefer weight-based, certain queries in compensatory environments.