An Enactive Approach to Facilitate Interactive Machine Learning for Co-Creative Agents

Abstract

This paper introduces a novel approach to developing co-creative agents that collaborate in real time creative contexts, such as art and pretend play. Our approach builds upon recent work in computational creativity called interactive machine learning (IML). In IML, agents learn through demonstration, interaction, and real time feedback from a human user (as opposed to offline training). To apply IML to open-ended creative collaboration, we developed an enactive model of creativity (EMC) based upon the cognitive science theories of enaction. This paper introduces our enactive approach to building co-creative agents within the broader field of interactive machine learning by describing the theory, design, and initial prototypes of two co-creative agents.