User Tools

Site Tools


theory_of_mind_in_robotics

Theory of Mind in Robotics

Part of Project Lirec

The ability to infer mental states in others that are not directly observable. Humans start to gain a ToM at around 3 years - i.e. a comprehension of mental states that can influence behaviour.

False belief task (Wimmer and Perner 1983)

In this experiment, a child is told the story of a boy called Maxi. In the presence of Maxi, Maxi's mother puts a bar of chocolate in a blue cupboard. Then, Maxi goes out to play, while his mother puts the chocolate in the green cupboard. When Maxi comes back and looks for the chocolate, the child is asked where Maxi will look after the chocolate: in the blue or in the green cupboard.

This test is usually solvable by children over 5 years old. After around the age of 10, theories of mind become fully developed (there is no difference in experimental testing with adults). This is demonstrated by the use of and understanding of phrases like:

  • “My mind wandered”
  • “My mind was racing”

Which indicate separation of the ideas of people and their mind's.

Primates have been proven to understand goals and intentions, but not desires and beliefs.

How ToM works

Two models:

  • Simulation theory: where we simulate other people's internal states, in order to understand them.
  • Theory Theory: we use exactly the same process as we do during examination of our own internal state (introspection).

Simulation theory has some basis in processes found in the human mind see Mirror Neuron.

Computational Models

PsychSim: Multi agent simulation - agents can have beliefs about other's beliefs.

Bosse, Memon and Treur: Mindreading and Emotion Reading.

ToM in Social robots

Cog: “Theory of Mind for a Humanoid Robot”, Scassellati.

theory_of_mind_in_robotics.txt · Last modified: 2009/01/13 12:56 by davegriffiths