Experimental Robot Psychology

摘要

In this paper I argue that an intentional methodology is appropriate in the design of robot agents in cooperative planning domains - at least in those domains that are sufficiently open-ended to require extensive reasoning about the environment (including other agents). That is, we should take seriously the notion that an agent's cognitive state expresses beliefs about the world, desires or goals to change the world, and intentions or plans that are likely to achieve these goals. In cooperative situations, reasoning about these cognitive structures is important for communication and problem-solving. How can we construct such models of agent cognition? Here I propose an approach that I call it experimental robot psychology because it involves formalizing and reasoning about the design of existing robot agents. It shows promise of yielding an efficient and general means of reasoning about cognitive states.