Abstract The epistemological tradition from Bacon to Kuhn produced a complete account of the context of justification — how knowledge claims are tested, extended, and revised — but no structural account of the context of discovery: the conditions under which the founding question of a new domain first becomes visible as a question. This article presents the Human-AI Dialectical Engine (HADE), a new architecture for systematic frontier discovery derived from two source methodologies — Method 1 (Deprivation as Genesis) and Method 2 (The Method of Frontier Knowledge) — and adapted for AI execution through the Artificial Deferral Suspension Protocol (ADSP) and the nine-phase formalisation procedure. The architecture’s central thesis is a precise division of epistemic labour: human researchers supply what AI structurally cannot — the pre-argumentative Phase 0 recognition that arises from lived formation under the Stripping Condition — while AI executes what human researchers cannot consistently achieve alone — the full nine-phase adversarial development of a frontier proto-claim to its Strongest Formulation, at the depth and breadth that rigorous frontier science requires. We show that this division is not a pragmatic convenience but a structural necessity derived from the Phenomenological Priority condition: genuine frontier recognition arrives as knowledge before proof, a pre-argumentative event that is irreducible to inference. We provide formal specification of the HADE’s five operational stages, demonstrate its application across 45 scientific and philosophical articles in twelve domains, and prove that the HADE’s AI component is subject to the IRM Impossibility Theorem — establishing that AI-assisted discovery systems are themselves SRSM systems whose value criteria undergo target drift without external human reference. The article closes by addressing the self-application question: whether a methodology for frontier discovery can itself be a frontier claim, and what constraints the Gödel-Turing-Lawvere family places on any formal methodology that attempts to certify its own validity. Keywords: frontier discovery; philosophy of science; AI epistemology; dialectical method; deprivation as genesis; human-AI collaboration; scientific methodology; self-reference; Phenomenological Priority; IRM Impossibility Theorem; context of discovery; ADSP; SRSM; Strongest Formulation; HADE
José Caetano de Mattos (Fri,) studied this question.