I will present a general theoretical framework for understanding quirky aspects of our beliefs and preferences, motivated by learning and evolutionary processes. This framework will use standard game theory tools, albeit with a corresponding reinterpretation of the key assumptions and results. To exemplify this approach, I will briefly present two toy models. The first model attempts to explain some of our non-consequentialist ethics, such as the omission-commission distinction, the means-byproduct distinction, "strategic ignorance," and inefficient giving. This model is based on a novel theoretical result relating "higher order beliefs" (what I believe you believe) to coordination games. In our second model, we will attempt to explain motivated reasoning--the fact that we are more likely to collect and convey (and internalize) information that supports our "desired" conclusion--by recourse to a new signaling model known as "evidence games" (a formalization of the optimal information conveyed in a persuasion context).
Moshe Hoffman is a Research Scientist & Lecturer at Harvard's Program for Evolutionary Dynamics.