Designing Socially Intelligent Agents For The Ultimatum Game
Peter M. Todd, Bernhard F.J. Borges · 1997
To build socially intelligent artificial agents, we must decide how much---and what kind of---intelligence to endow them with. Vriend (1997) has recently questioned whether adding reasoning will help or hinder the behavior of simple learning agents in social games. We show that adding the right kind of reasoning will help agents act more intelligently by learning more quickly. Traversing the Spectrum of Reasoning Ability How much intelligence does it take to be socially intelligent ? Some researchers have proposed that social intelligence involves the rather complex ability to model the intentions and goals of the agents with whom one interacts socially (Byrne & Whiten, 1997). Others have suggested that simpler, " fast and frugal" mental heuristics may serve us in the social realm (Gigerenzer, 1997). Between these two points---high intelligence in the first case, and limited cognition in the second---lies a large range of possibilities. If we are interested in actually buildin...