Neither Tool nor Collaborator: Rethinking Human–AI Co-Creativity in Artistic Practice with Material Engagement Theory

Paper Title

Neither Tool nor Collaborator: Rethinking Human–AI Co-Creativity in Artistic Practice with Material Engagement Theory

Publication Info

• Topic area: Human–AI co-creativity in artistic practice
• Keywords: Human–AI co-creativity, Material Engagement Theory, machine learning systems, artistic practice, aesthetic experience, creative affordances, meta-artist, iterative engagement, relational dynamics

Background and Problem

• Problem / challenge: Existing frameworks for human–AI co-creativity polarize AI systems as either passive tools or autonomous collaborators, failing to capture the relational and emergent dynamics of artistic practice with machine learning systems.
• Significance: Understanding human–AI co-creativity as relational and emergent can enhance artistic practices, inform the design of digital creativity support systems, and reshape conceptualizations of creative agency.
• Motivation and related work: Prior studies have framed AI systems as either tools or collaborators, limiting the exploration of emergent creative and aesthetic affordances. Material Engagement Theory (MET), originally developed in cognitive archaeology, offers a relational perspective that dissolves the divide between artist and material, enabling a deeper understanding of human–AI co-creativity.

Solution

• Proposed approach: Application of Material Engagement Theory (MET) to reframe human–AI co-creativity as emergent and relational, focusing on sustained interactions between artists and machine learning systems.
• Novelty:
  1. Recasting human–AI co-creativity through MET to overcome dualistic framings of AI systems.
  2. Identifying mechanisms such as iterative engagement along a control–chance continuum, artistic intuition, internalization of system affordances, and the evolving role of the meta-artist.
  3. Highlighting emergent creative and aesthetic processes shaped by reciprocal engagement.
  4. Providing implications for HCI, including design recommendations for creativity support systems and intent-based interactions.
• Procedure and key techniques:
  • Qualitative, interpretive, and framework-guided analysis of 18 contemporary artists’ practices using machine learning systems.
  • Integration of large language model (LLM)-assisted information extraction with manual thematic analysis.
  • Examination of 54 publicly available documents to identify MET concepts and emergent mechanisms.

Results

• Concrete findings:
  • Artists engage with machine learning systems through iterative cycles of generation and feedback, fostering mutual influence and emergent creative and aesthetic affordances.
  • Mechanisms such as artistic intuition, internalization of system logic, and the meta-artist role were identified as central to relational co-creativity.
  • Aesthetic significance arises through interpretive framing, emotional filtering, and viewer engagement with ambiguous outputs.
• Advantage over baselines:
  • MET provides a unified relational framework that transcends the dualistic framing of AI systems as tools or collaborators, capturing the emergent dynamics of human–AI co-creativity.
• Experiments / evaluation:
  • Analysis of 18 artists’ practices using thematic coding of validated data extracted from interviews, essays, and presentations.
  • Manual validation of LLM-assisted extractions ensured accuracy and completeness, with a recall rate of 90%.
• Limitations and future work:
  • Constraints of MET in addressing the abstract nature of digital materiality and mediated human–AI interactions.
  • Limited generalizability to non-visual artistic domains.
  • Future research could explore mechanisms like intuition and internalization in greater depth and extend findings to other creative contexts.

Summary

This study rethinks human–AI co-creativity using Material Engagement Theory (MET), framing artistic practice with machine learning systems as relational and emergent. Key mechanisms such as iterative engagement along a control–chance continuum, artistic intuition, internalization of system affordances, and the meta-artist role were identified. Findings highlight how creative and aesthetic properties arise through sustained interaction, offering novel insights for HCI design and digital creativity support systems. The study contributes to understanding co-creativity as a dynamic interplay between human intention and machine responsiveness, opening avenues for future research and practical applications.