AI Drawing Partner: Co-Creative Drawing Agent and Research Platform to Model Co-Creation

TL;DR

The paper addresses the need to quantify co-creative AI interactions by introducing the Co-Creative Sense-Making (CCSM) framework and the AI Drawing Partner, a real-time co-creative drawing agent and data-collection platform. It demonstrates CCSM in a ten-session case study, showing how cognitive, interaction, collaboration, and domain dynamics can be automatically coded and visualized to reveal differences between abstract and representational drawing strategies. The primary contribution is the CCSM framework and the AI Drawing Partner as a domain-independent research platform; the secondary contribution is a methodology for automatic coding and visualization of co-creative processes, with potential cross-domain applicability. The work advances human-centered AI by enabling standardized, scalable analysis of co-creative processes, offering insights for design, education, and therapy, and providing a publicly accessible platform for ongoing research.

Abstract

This paper describes the AI Drawing Partner, which is a co-creative drawing agent that also serves as a research platform to model co-creation. The AI Drawing Partner is an early example of a quantified co-creative AI system that automatically models the co-creation that happens on the system. The method the system uses to capture this data is based on a new cognitive science framework called co-creative sense-making (CCSM). The CCSM is based on the cognitive theory of enaction, which describes how meaning emerges through interaction with the environment and other people in that environment in a process of sense-making. The CCSM quantifies elements of interaction dynamics to identify sense-making patterns and interaction trends. This paper describes a new technique for modeling the interaction and collaboration dynamics of co-creative AI systems with the co-creative sense-making (CCSM) framework. A case study is conducted of ten co-creative drawing sessions between a human user and the co-creative agent. The analysis includes showing the artworks produced, the quantified data from the AI Drawing Partner, the curves describing interaction dynamics, and a visualization of interaction trend sequences. The primary contribution of this paper is presenting the AI Drawing Partner, which is a unique co-creative AI system and research platform that collaborates with the user in addition to quantifying, modeling, and visualizing the co-creative process using the CCSM framework.

Figures (12)

• Figure 1: Raw coded values (left) and creative sense-making curve (right) for a five-minute co-creative drawing session.
• Figure 2: The Co-Creative Sense-Making Framework. Each category has a number of features that can be quantified. This categorization serves as a data collection schema for quantified co-creative AI systems.
• Figure 3: Action history overlaid onto the CSM curve for the user in Session 6.
• Figure 4: AI Drawing Partner interface with annotated functionality.
• Figure 5: The AI-Human communication channels in the AI Drawing Partner.