UC San Diego
Essays in decision making under cognitive load
- Author(s): Sanjurjo, Adam Angel
- et al.
My first chapter tests several hypotheses of information overload in an experiment where subjects estimate security prices under variable signal loads. Peter Katuscak and I find that as information load increases subjects eventually stop assimilating further information, and they shift increasing weight towards the most salient information. This combination of results leaves information receivers vulnerable to strategic manipulation by senders. My second chapter builds on the multiple attribute search experiment, and analysis, of Gabaix and Laibson (2006) in four ways; I provide a basic description of subjects' search behavior, study behavior on the individual subject level, provide a partial characterization of optimality, and compare subjects' behavior to my partial characterization of optimality. I find that subjects' search behavior violates optimality at high rates, but is also highly systematic, that 98% of all search behavior can be explained by four simple, exclusive, types, and that subjects often search conditionally too deeply within alternatives and exhibit strong adjacency biases in switching between alternatives. I also observe unambiguous evidence of memory failure by subjects. My third chapter tests the hypothesis that working memory limits can explain subjects' main systematic deviations from optimality, as well as other fact patterns in search, in the Gabaix and Laibson (2006) dataset. First I show that the most popular type of search pattern by subjects also requires the unique minimum amount of working memory load. Second, I show that more systematic search sequences require less working memory load than more "random- looking" sequences. These theoretical results strongly suggest that a simple model of search in which working memory is limited, but subjects otherwise search optimally, can explain Gabaix and Laibson's subjects' main systematic deviations from optimality as well as other fact patterns from the dataset