A Systematic Review of Artistic Human-AI Co-Creative Systems and Interaction Design

Abstract

The co-creativity community is forging ahead with building more sophisticated and tailored systems to support and enhance human creativity. Design considerations or learnings from previous work can be an important and efficient starting point for this. To assist with that, we created a systematic literature review of 63 papers on artistic co-creative systems that clarifies what current co-creative systems actually do, how they support creativity, and qualitative experiences of what users value or struggle with when working with them. The contribution of the paper also includes a set of 15 actionable design considerations that link system design choices to their impact on creativity and user experience. The selected papers span a diverse range of artistic applications, from visual arts to design, writing, etc, where the AI's role goes beyond that of a tool- to engaging with the creative process or the user as an active collaborator. From the literature surveyed, we identified several dimensions of relevance to the design of such systems- (1) Phase of the creative process, (2) Creative task, (3) Proactive behaviour of the system, (4) User control, (5) System embodiment, (6) AI Model type. We see repeated tendencies across systems, such as users responding differently to the same behaviour depending on how the system's role is framed, and shared initiative leading to higher satisfaction and smoother collaboration when the balance is slightly tipped toward the human; these are simply two examples among many patterns identified across the six dimensions, of which only a small subset can be included in the abstract. Despite advancements, key limitations persist, including a lack of systems addressing early creative phases like problem clarification, and the complexity of user adaptation to AI tools.