An Agent for the Prospect Presentation Problem

Evaluating complex propositions that are composed of several lotteries is a difficult task for humans. Presentation styles can affect the acceptance rate of such proposals. We introduce an agent that chooses between two presentation methods, while aspiring to maximize proposal acceptance. Our agent uses decision theory in order to model human behavior and uses the model to select the presentation which maximizes its expected outcome. We examine several decision theories, and use machine learning to adapt them to our domain. We perform an extensive evaluation of our agent in comparison to other baseline agents and show that presentation can indeed affect the acceptance rate of propositions and that the agent we propose succeeds in selecting beneficial presentations.