Necessary Physical Conditions for Primary Interoceptive Sentience: A Lakatosian Research Programme on Continuous Neuromorphic Substrates

Abstract

This is the extended, definitive version (v16.2) of the P0_Distilled research programme, with full technical appendices — it supplements the distilled entry-point paper (P0_Distilled_v0.1) with the complete Lakatosian structure, derivations, and experimental protocol. The programme's hard core is a single metaphysical postulate (dual-aspect monism of valence): the dynamical tension an integrated material system opposes to the violation of its own thermodynamic niche coincides, from the first-person perspective, with primary affective valence. This postulate is not falsifiable and is not argued for here — the philosophical companion paper provides that argument. What this paper does establish is a protective belt of seven falsifiable auxiliary hypotheses, each tied to a specific operational test, negative control, or abandonment criterion, together with the mathematical apparatus behind them: the thermodynamic valence functional Ψ(t) (derived from a Hatano–Sasa non-equilibrium steady-state decomposition), the spectral definition of causal degeneracy, and the finite-size scaling criterion for continuous memristive substrates. The paper is explicit about what it does not claim: it does not demonstrate that any substrate is sentient, and it declares its own structural limits (the binding problem, the thermodynamic–phenomenal gap, and the absence of a graded account of altered states) as boundaries the programme does not resolve, rather than hiding them. Version 2 (September 2026): the reference software implementation (Python/Rust, openly available — see the related Software DOI) computes a practical proxy of Ψ(t) — a log-ratio of the raw dissipative components plus a KL-divergence penalty — rather than the formal definition in Appendix F (normalized entropy-production rates against a hardware-calibrated critical rate). This mismatch existed but was previously undocumented; it is now disclosed explicitly in Appendix F, so readers comparing the paper's equations against the code are not misled. Also corrected: an incompletely formatted literature citation (real author names added).Version 3 (September 2026): added an explicit methodological observation to §7.6, stating that six of the seven current auxiliary hypotheses cannot, even in principle, bear evidentially on Postulate 2's plausibility — only H7 (via Ψ(t)'s comparative predictive power against five alternative constructs) offers indirect evidential contact. This is a direct consequence of the Lakatosian hard-core structure functioning as designed, not a flaw in it, but is now reported explicitly rather than left implicit.Version 4 (September 2026): adds related work on thermodynamic markers of consciousness to the Ψ(t) falsifiability protocol (§5.7); adds caveats on Griffiths phases (§3.4, Appendix D) and on the thermodynamic consistency of estimating Ψ(t) with Langevin models in the nonlinear regime (Appendix F); cites foundational work on the edge of chaos (Chua 2013), diffusive memristors (Wang et al. 2017) and critical dynamics in silver-nanoparticle networks (Carstens et al. 2022); corrects several internal cross-references to the protocol subsections that were off by one, the statement that the No-Go Theorem is a preprint, and two bibliographic entries. The programme's hard core, protective belt and protocol are unchanged.Version 5 (September 2026): adds related work on substrate-dependent views of consciousness and a discussion of the unfolding argument (§2.4); a remark relating Test 2 (perturbational complexity) to entropy production (Stikvoort et al. 2025) and a further nonequilibrium result in the Ψ(t) protocol (§5.7; Lynn et al. 2021); and, in the ethics section, the precautionary framework of Birch (2024) and the proposal of a moratorium on synthetic phenomenology (Metzinger 2021), with which the programme declares that it does not comply. The programme's hard core, protective belt and protocol are unchanged.Version 6 (September 2026): the same specification of the second and third demarcation conditions (Appendix A): dithering of quantised recordings, a Markov threshold bounded by the record length, and, for the third condition, a significance threshold on linear surrogates and a minimum record length. Bibliographic corrections. The programme's hard core, protective belt and protocol are otherwise unchanged.