• Summary
• Related Documents
  • Active Bibliography
  • Co-citation
• Version History

Abstract:

If we are to build human-like robots that can interact naturally with people, our robots must know not only about the properties of objects but also the properties of animate agents in the world. One of the fundamental social skills for humans is the attribution of beliefs, goals, and desires to other people. This set of skills has often been called a \theory of mind." I propose to implement a basic theory of mind on a humanoid robot. The implementation will be based on Leslie's [36] and Baron-Cohen's [3] models of the development of theory of mind in human children. Following Leslie's decomposition, the robot will distinguish between animate and inanimate objects on the basis of self-propelled motion. Objects that violate an \intuitive physics" similar to Leslie's ToBY (theory of body) module will be treated as animate. Animate objects will then be further processed by the \intuitive psychology" systems in Baron-Cohen's model. I will implement the rst three modules of Baron-Cohen's model: an intentionality detector (ID) which attributes basic goals and intent to animate objects based on their motion patterns, an eye direction detector (EDD) which determines the object that a person is looking at, and a shared attention module (SAM) which directs behavior to attend to an object of mutual interest. The implementation of these systems will allow for a potential analysis of the human models that serve as a starting point. Using these systems, the humanoid robot will be able to act based on the perceived intent of an individual. For example, when a person stands in front of the robot and turns to look at an object in the robot's workspace, the robot will oer the desired object to the person. 1

Citations