The Empirical Programme: Four Tasks

What this section is

The consciousness chapters state what UHM proves about the coherence matrix $\Gamma$ and what it only interprets. This section collects, in one place, the work that proofs cannot do: fixing which coherence corresponds to which quality, checking that the geometry of experience is the geometry the theory commits to, and building systems in which the mechanism runs. Nothing here is new doctrine. The pages gather protocols and criteria scattered over the corpus, give each claim the status of its registry row, and mark every new proposal as a research programme [Pr] or a hypothesis [H].

The starting point

UHM's position on experience has three parts, and they carry different statuses.

1. The identity. $\Gamma$ is one object with an external side (physics) and an internal side (experience). This is an ontological position, two-aspect monism, with status [I]: it "reformulates the hard problem rather than solving it" (epistemic status).
2. The mathematics. What is proved about the internal side is proved as mathematics: a viable dissipative holon has $\mathrm{Coh}_E > 1/7$ (T-38a [T]; the identification "E = interiority" is a postulate [P] and the reading "no zombies" is [I], registry row 38a); the phenomenal functor is unique under the axioms (uniqueness [T]); a quality is fixed exactly by its dissimilarities to all others, and to within twice the covering radius by finitely many probes (enriched Yoneda for qualia [T], the construction of Tsuchiya, Phillips & Saigo 2022 applied to UHM's quality space).
3. The residue. Two things are left open on purpose. Why this structure is felt is not explained (what UHM does not explain, item 1), and T-214 [T] shows that the bridge from states to experience cannot be an internal morphism of the theory. Which ray $[\lvert q\rangle]$ is red is "an empirical question, analogous to determining the mass of the electron" (item 2; the earlier identification of this residue with the $G_2$-frame is retracted).

The second residue is where proofs stop and measurement starts. That is why this section exists.

Four tasks and what decides them

Task	Question	What decides it	Kind of evidence	Page
(i) Calibration	Which coherence pattern corresponds to which quality, and how a substrate maps into $\Gamma$	Data: reports, discriminations, similarity judgements, psychophysics, neural signals	Third-person measurement against a declared ground truth	Calibration
(ii) Structure	Is the geometry of quality space the geometry $\Gamma$ commits to: seven axes, Fano relations, the window $2/7 < P \leq 3/7$, the F-Band sums	Data: similarity structures, reconstructed $\Gamma$, their invariants	Third-person measurement of relations, not of single qualities	Structure
(iii) Mechanism	Can a system be built in which $\Gamma$-dynamics, regeneration and a self-model in the window actually run, and do the predicted functional signatures appear and disappear with them	Engineering of a cognitive architecture, with ablation tests	Construction plus third-person tests on the construction	Engineering
(iv) The hard problem	Why any of this is felt at all	Neither data nor engineering	None possible: every datum is third-person	This page, below

(i) Calibration is empirical

The mathematics fixes the form of the correspondence — the functor $F$ into rays of $\mathbb{P}(\mathcal{H}_E)$ with the Fubini–Study metric — but not its constants. Two calibrations are needed, and neither can be derived:

• the measurement calibration: the free parameters $\theta$ of a reconstruction $\pi_{\mathrm{bio}}$ (or of the map $G$ for an artificial system) that takes signals to $\hat\Gamma$ (measurement protocol);
• the phenomenal calibration: which ray is which quality, and the monotone map $f$ from perceived dissimilarity to $d_{\mathrm{FS}}$ that the enriched Yoneda theorem needs before it can be tested ("the refutation needs a calibration $f$ … which is not fixed").

Both are settled by fitting to data where the answer is independently known, and by testing on data that did not enter the fit. The corpus has one hard rule for this: no threshold without ground truth.

(ii) Structure is empirical, and it is the strongest channel

A single calibrated quality supports nothing: any theory can be fitted to one point. A relation among many qualities is different. UHM commits in advance to a geometry — complex projective space with the Fubini–Study metric for quality, seven axes and seven Fano lines for $\Gamma$, a window and two bands for a living state — and that commitment forbids specific patterns (for example, dissimilarity matrices that no configuration of rays can realise). A theory that has fixed its geometry before the data can lose to the data. This is why the structural level is the strongest empirical support the identity can receive: it cannot prove the identity [I], but a structure that survives many chances to fail is what makes the identity worth holding. The structure page lists each invariant with its status and what would confirm or refute it.

(iii) The mechanism is an engineering task

The theory specifies a mechanism: CPTP dynamics of $\Gamma$, regeneration $\mathcal{R} = \kappa(\Gamma)(\varphi(\Gamma) - \Gamma)\,g_V$ read through a self-model, viability in the window. Whether that mechanism can be realised, and what a realisation does, is answered by building one and testing it — including ablations that remove the self-model, the $E$-coherences or a Fano line and check that the predicted signatures go with them. This is the programme of the engineering page, which links the corpus's AI chapter, the ten in-silico tests, and the Phase I of the experimental protocol.

(iv) The hard problem is decided by neither

Chalmers' question — why physical processing is accompanied by experience at all — is not a calibration problem and not an engineering problem.

• Not by data. Every measurement — a report, a discrimination, a similarity judgement, an EEG trace, a reconstructed $\hat\Gamma$ — is a third-person record. A perfect correlation between records and $\Gamma$ still leaves open why the correlated structure is felt. Inside UHM this is sharpened, not lamented: the environment's noise channel carries exactly zero information about the phase-carried quality (T-302 [T]), and no bridge from states to experience is an internal morphism (T-214 [T]).
• Not by engineering. A system that passes every functional test shows that the mechanism runs. Whether it is felt is again the identity, not an extra test result.
• What UHM does with it. It closes the question by identity — the internal side of $\Gamma$ is experience [I] — and says plainly that this is a position, "the boundary between map and territory" (two-aspect monism), not a derivation. This residue is not specific to UHM [I]: every theory of consciousness ends with one unexplained primitive — a psychophysical law, an identity, a fundamental property — and the comparison of axiomatic choices shows that UHM's is one primitive, not two or three.

The practical consequence is a division of labour. The hard problem is acknowledged once, here, and not re-argued on the other three pages; they deal only with what data and engineering can decide.

What this section does not claim

• It reports no data. Every prediction collected here is untested unless its source page says otherwise (see the verdict table of the falsifiability page).
• It does not upgrade any status. Proposals made here for the first time are [Pr] or [H]; a protocol that passes would corroborate a claim at its existing status, and one that fails would refute it at that status.
• It does not read consciousness off a measurement in systems where no ground truth exists — language models, fungal networks, simulated agents — for the reason given in the substitution theorem.
• It does not revive the withdrawn numerical bridge between the perturbational complexity index and purity: "the closeness of $\mathrm{PCI}^* = 0.31$ to $2/7 \approx 0.286$ is a coincidence of two unrelated scales" (Step 5); the testable form is a concordance of verdicts (P8.4, SUB-5).

Sources gathered in this section

Source	What is taken from it
Two-aspect monism	The identity [I], the residues, relational definiteness, the Tsuchiya–Saigo precedent
Categorical formalism §6.2.1	Enriched Yoneda for qualia [T]: isometry, finite probes, probe counts, realisability
Qualia structure	21 channels, 7 holonomies, the decoder T-301, fading, access conditions
Emotional taxonomy	Valence and arousal from $dP/d\tau$
Falsifiability	Predictions 1–4, refutation criterion, F-Gap, F-ISF, F-Neural, F-Band
Measurement protocol	$\pi_{\mathrm{bio}}$, P8.1–P8.6, the substitution theorem, SUB-1…SUB-6
Experimental protocol	Four phases, Γ-native agent, three-level falsification
CC predictions	23 predictions, decision protocols, falsification levels
Phenomenology map	Experience → observable → test
AI consciousness	Architectural requirements, tests E1–E10
Console roadmap	Validation stages V0–V2 for a self-report and wearable instrument

Next: Calibration →