As artificial intelligence evolves from a passive tool into a collaborative actor with autonomous decision−making and learning capabilities, Human−Computer Interaction (HCI) is undergoing a paradigm shift from "humans using technology" toward Human−AI Collaboration. The growing agency of AI challenges the foundational assumption of traditional HCI that humans are the sole agentive actors. However, the theoretical premise of user−centered design struggles to accommodate dynamic multi−agent collaboration, and existing evaluation frameworks lack effective instruments for assessing collaboration quality and equitable value distribution. These limitations result in the absence of collaborative frameworks for heterogeneous actors, a lack of mediating tools and controllable experimental spaces for cognitive alignment, and the difficulty of institutionalizing localized collaborative experience into public capacity. Grounded in socio−technical systems theory, this paper proposes a four−element model—Heterogeneous Communities, Provocative Prototypes, Peripheral Innovation, and Collaborative Networks—reveals the cyclic evolutionary mechanism among these elements, and provides differentiated recommendations for researchers, research managers, and policy makers, offering a theoretical framework and practical pathway for building an inclusive, equitable, and sustainably evolving Human−AI collaborative system.

human−AI collaboration  /  human−computer interaction  /  innovation ecosystem  /  socio−technical systems  /  paradigm transformation