Human-Robot Interaction in Concept Acquisition: a computational model

Abstract

This paper presents a discussion and simulation results which support the case for interaction during the acquisition of conceptual knowledge. Taking a developmental perspective, we first review a number of relevant insights on word-meaning acquisition in young children and specifically focus on concept learning supported by linguistic input. We present a computational model implementing a number of acquisition strategies, which enable a learning agent to actively steer the learning process. This is contrasted to a one way learning method, where the learner does not actively influence the learning experience. We present results demonstrating how dyadic interaction between a teacher and learner may result in a better acquisition of concepts.