In this article, an enactive architecture is described that allows a humanoidrobot to learn to compose simple actions into turn-taking behaviors whileplaying interaction games with a human partner. The robot's action choices arereinforced by social feedback from the human in the form of visual attentionand measures of behavioral synchronization. We demonstrate that the system canacquire and switch between behaviors learned through interaction based onsocial feedback from the human partner. The role of reinforcement based on ashort term memory of the interaction is experimentally investigated. Resultsindicate that feedback based only on the immediate state is insufficient tolearn certain turn-taking behaviors. Therefore some history of the interactionmust be considered in the acquisition of turn-taking, which can be efficientlyhandled through the use of short term memory.
展开▼
