The Role of Signaling Action Tendencies in Conflict Resolution

We investigate various strategies for stopping games embedded in the larger context of an artificial life simulation, where agents compete for food in order to survive and have offspring. In particular, we examine the utility of letting agents display their action tendencies (e.g., "continue to play" vs. "quitting the game" at any given point in the game), which agents can take into account when making their decisions. We set up a formal framework for analyzing these "embedded stopping games" and present results from several simulation studies with different kinds of agents. Our results indicate that while making use of action tendency cues is generally beneficial, there are situations in which agents using stochastic decision mechanisms perform better than agents whose decisions are completely determined by their own and their opponents' displayed tendencies, particularly when competing with agents who lie about their action tendencies.