日本語タイトル:共創知性:人間とAIによる理論生成プロセスの新しいモデル Abstract:Artificial intelligence systems are no longer confined to roles such as data processing, automation, or post hoc analysis. Increasingly, they participate directly in the generative phases of human thought—hypothesis formation, conceptual exploration, and structural refinement. Despite this shift, dominant academic frameworks continue to treat AI as an external tool: a passive instrument operated by an active human subject. In practice, however, many researchers already engage in sustained, iterative dialogue with AI systems as cognitive partners. Yet the structure of this joint theoretical thinking process remains largely unexamined. This paper introduces the concept of Co-Creative Intelligence, defined as a mode of theoretical discovery in which novel conceptual structures emerge from sustained, structured interaction between human cognition and artificial intelligence systems. We argue that the absence of a formal model for human–AI joint theoretical discovery constitutes a critical gap in contemporary cognitive science, artificial intelligence research, and meta-scientific inquiry. To address this gap, we propose a framework that models theoretical discovery as a distributed cognitive process enacted across heterogeneous agents within a shared informational substrate. This framework characterizes the dynamic transitions of representational control between human and AI cognitive layers, as well as the role of lateral transitions in enabling non-linear restructuring across conceptual regions. Importantly, the present study does not aim to validate the correctness of the STM Layer Theory itself. Rather, the theory is treated as an empirical trace of a co-creative process and is analyzed as a case through which the structure of human–AI theoretical generation can be described. In this sense, the apparent circularity between theory and meta-theory is not a logical flaw but a constitutive feature of the phenomenon under investigation. By formalizing the dynamics of human–AI co-creation, this work contributes a foundational model for understanding how theoretical knowledge may be generated in the emerging landscape of AI-augmented research.
STM (Wed,) studied this question.