AI as an Epistemic Amplifier: A Recursive Theory of Conceptual Origination and Exploration

Recent advances in artificial intelligence have accelerated scientific discovery, yet the epistemic mechanism underlying this acceleration remains poorly understood. This paper proposes that AI functions as an epistemic amplifier within a recursive feedback loop linking conceptual origination and conceptual exploration. Drawing on Boden’s (1998, 2004) typology of computational creativity, Floridi’s (2011) philosophy of information, Dretske’s (1981) information-theoretic epistemology, Merleau-Ponty’s (1945/2012) phenomenology of perception, and empirical developments including AlphaProof (Hubert et al., 2025), AlphaFold (Jumper et al., 2021), and RFdiffusion (Watson et al., 2023), this paper formalizes the Principle of Recursive Origination–Exploration Amplification (ROEA). ROEA holds that origination introduces new epistemic structures enabling exploration, and exploration recursively amplifies subsequent origination through three mechanisms: anomaly amplification, combinatorial amplification, and cognitive liberation. The principle specifies conditions under which amplification succeeds or fails, positioning it as a structural-descriptive account of epistemic acceleration in human–AI cognitive systems.