Title: Interaction-Centered Intelligence: Positioning Interaction as the Core Unit of Analysis in Human-AI and Co-Creative Systems

Abstract:

Conventional artificial intelligence frameworks have predominantly defined intelligence as discrete, isolated computation confined within self-contained agents. In fields ranging from classical AI and machine learning to contemporary generative models, the standard metric for assessment remains the individual model or autonomous system, judged by its outputs, benchmark scores, predictive precision, or optimization efficiency. Although these methodologies have driven significant progress, they frequently overlook the critical function of interaction in fostering intelligence, creativity, meaning-making, and adaptive behaviors.

This study advocates for interaction to serve as the fundamental unit of analysis for co-creative AI and the broader domain of interaction-centered intelligence. By integrating perspectives from distributed and embodied cognition, enaction, participatory sense-making, human-computer interaction, and computational creativity, the paper outlines a historical shift toward increasingly relational definitions of intelligence. Leveraging previous research in Creative Sense-Making, quantified co-creation, and co-creative platforms like the Drawing Apprentice and AI Drawing Partner, the argument posits that intelligence arises from dynamic interactions among agents, their environments, and socio-technical systems, rather than being confined to internal computational processes.

The paper presents "Interaction-Centered Intelligence" as a theoretical framework designed to elucidate human-AI co-creation, collaborative emergence, adaptive participation, and interactional dynamics. Moving beyond an exclusive focus on generated outputs, this framework prioritizes the examination of interaction trajectories, coordination patterns, participatory engagement, adaptive regulation, and the phenomenon of interactional drift as they unfold over time. The discussion concludes with the implications of this approach for explainable co-creative AI, hybrid intelligence, enactive AI, and the design of future human-AI systems.

Source: arXiv Generated at: 2026-06-02 00:00:00 UTC