A Framework for Dialogue-Based Human-AI Creative Collaboration

Abstract

Human-AI co-creative collaboration has been proposed as a model that integrates the strengths of both humans and creative algorithms. Several frameworks have been developed to classify and guide the design of such systems. However, these models lack communication mechanisms that enable the emergence of a common ground between humans and machines through a mutual adaptation of understanding about goals and meanings, a crucial component in all collaborations. We argue that dialogue is a mechanism that serves this purpose and can be included in human-AI co-creative systems to that end. We propose a breakdown of dialogic creative interaction and use it to analyze co-creative dialogue with GPT-3.