Interaction-Centered Intelligence: Toward an Interaction-Based Theory of Human-AI Co-Creation

The paper proposes a novel framework called Interaction-Centered Intelligence, positioning interaction as the primary unit of analysis for human-AI co-creation. It critiques traditional models that focus on isolated computation and benchmarks, advocating for a relational understanding of intelligence that emphasizes interaction dynamics among agents and environments. This approach has significant implications for the design of explainable co-creative AI systems, as it shifts the focus from output evaluation to understanding interaction trajectories and adaptive participation in collaborative contexts.