If internal theories of consciousness face inherent limits, what do you do?
You stop trying to prove what consciousness is from the inside. Instead, you specify what it does from the outside, formalize the mechanism, derive predictions, and let experiments decide.
The framework advances three nested claims at different strengths. Level A says what consciousness does. Level B says how. Level C says what kind of reality that implies.
Level A (functional) [A]: Consciousness selects among possible futures. It is a navigational faculty that reduces uncertainty over trajectory distributions. This claim is equivalent to active inference at the classical level; its contribution is architectural rather than empirical, revealing IIT, GNWT, and active inference as subsystems of a single navigational faculty. Level A is compatible with classical implementations and does not require exotic physics.
Level B (mechanistic) [B]: That navigational selection depends on quantum probability sculpting via radical pair spin coherence (a type of quantum behavior in paired molecules), amplified by criticality (the brain's tendency to operate at the tipping point between order and chaos) and adaptive measurement basis selection (the organism's ability to choose what aspect of the quantum system to interact with). Each of these is explained in Chapters 8 and 9. This is the idea that stops people mid-sentence. A living system, through quantum interference in radical pair spin states, can reshape the probability landscape of its own future. Not compute faster. Not process more information. Sculpt the odds themselves, in ways that classical systems provably cannot.
Level C (ontological) [C]: That possibility space is a physically real structure and consciousness physically participates in the actualization of definite outcomes from genuine indeterminacy. This is the most speculative claim. It is motivated by the evolutionary argument and the analysis of temporal experience but is not required for Levels A or B.
Each level can be tested independently, and each can fail without bringing the others down.
Karl Friston's Free Energy Principle and its active inference extension frame consciousness as minimizing surprise: continuously generating models of the world, updating them when predictions fail. Action selection decomposes into epistemic terms (reduce uncertainty) and instrumental terms (fulfill preferences).
Level A is active inference, described from the outside. "Consciousness navigates possibility space" and "consciousness minimizes expected free energy" are two descriptions of the same process. One from the navigational perspective, one from the information-theoretic perspective. This is shared ground, not contested territory.
So what does Level A actually contribute? Architecture, not new predictions. Active inference describes what an individual agent does. IIT describes what the substrate must be. GNWT describes how the substrate communicates. These three frameworks developed largely independently and their relationship to each other stayed unclear. The four-operations framework (Generation, Selection, Integration, Propagation) shows they are subsystems of a single navigational architecture. IIT's integration is what makes navigation coherent. GNWT's broadcasting is how the navigational output reaches the rest of the brain. Active inference's surprise minimization is what navigation accomplishes. They were never competitors. They were describing different parts of the same faculty.
That unification is Level A's contribution. It doesn't generate novel predictions beyond what each component theory already generates. Its value is revealing the structural relationship and providing the architectural frame within which Level B's quantum mechanism makes sense. Level B is where NFT departs from active inference, and it departs significantly: navigation's generative mechanism involves quantum probability sculpting, which provides capabilities that classical Bayesian inference cannot replicate.
Any theory worth taking seriously should tell you, in advance, exactly what would prove it wrong. This is not a concession. It is the price of admission to science. Here is what would kill NFT.
Level B dies if:
- The microtubule substrate is conclusively too decohered for any quantum effects relevant to spin coherence. (The quantum parts of neurons turn out to be too noisy for the mechanism to work.)
- Classical neuromorphic systems at criticality are behaviorally indistinguishable from quantum neuromorphic systems on the same hardware. (You can build the same navigation without quantum effects, meaning the quantum part was never necessary.)
- The ENAQT calculation shows no systematic bias at physiological parameters. (The math says the quantum advantage vanishes at body temperature.)
- No radical pair signatures are detectable in neural tissue under physiological conditions. (The chemistry simply doesn't happen in living brains.)
Level A dies if:
- Consciousness is shown to have no measurable effect on trajectory entropy reduction beyond what unconscious processing achieves. (Conscious and unconscious systems navigate equally well, meaning consciousness isn't doing what we say it does.)
- The four-operations framework fails to unify IIT, GNWT, and active inference. (One of these theories turns out to be incompatible with the navigational architecture rather than a subsystem of it.)
Level C dies if:
- The philosophical and physical arguments for the block universe are decisively overturned. (This is primarily a philosophical/interpretive question rather than a straightforward empirical one. The eternalism/presentism debate is unlikely to be settled by a single experiment. But if the physics community converges on an interpretation of quantum gravity that is incompatible with the block universe, Level C loses its ontological foundation.)
- No quantum measurement effects are detectable at biological scales. (The physical mechanism through which consciousness would participate in actualization doesn't exist.)
Where we are. Every leading theory of consciousness captures something real but fails to capture everything. The pattern of partial success may be structural, not temporary. NFT responds by defining consciousness from the outside: what it does (navigate), how it does it (quantum probability sculpting), and what it navigates through (possibility space). Three layers, each independently testable, each independently killable. That is the architecture. Whether it holds weight is what the rest of this book is for.