Most responses to the hard problem are defensive maneuvers. Global Workspace Theory explains access consciousness and hopes nobody notices the residue. Integrated Information Theory quantifies something and calls it experience. Higher-Order Thought theories add a meta-representational layer and trust that phenomenality will follow. Even physicalists who think the gap will close eventually have already conceded Chalmers his framing — yes, there is an explanatory gap; it will simply yield to better neuroscience. That concession is the place the field has spent thirty years.
Carlo Rovelli refuses the concession. So does “Beyond the Bat.” Neither argues that the gap is narrowing. Both argue that the gap was never there to begin with. This is dissolutionism in the strict sense, and it is rare enough that the agreement deserves close attention — especially because Rovelli and RSP reach the same destination by routes so different that understanding their relationship illuminates what dissolution actually requires.
Rovelli’s essay is short, clean, and devastating to the standard framing. Its core move is epistemological. The explanatory gap, he argues, is an artifact of a methodological error: treating scientific knowledge as a view from nowhere and then discovering, unsurprisingly, that the first-person perspective cannot be located within it. There is no such view. Knowledge is always embodied, perspectival, realized within the very reality it describes. Mind is brain behavior described at a high level — obtained by subtraction from a complete physical account rather than by addition of something extra. Qualia are names for processes. Zombies are self-defeating. The soul is real, and it is part of nature.
I agree with all of it. I also think it is not enough.
What Rovelli Gets Right
The agreements are not throat-clearing. They form the backbone of any serious dissolution. Four of them are load-bearing.
The view from nowhere does not exist
Rovelli’s most important sentence is this: “The world is real, but any account of it can exist only from within it. Any knowledge is perspectival.” The mistake at the root of the hard problem is treating the third-person scientific picture as a description from outside the world it describes, and then asking how the first-person could possibly fit inside that picture. The question is unanswerable because it is malformed. Any account is already from within. The dualism we discover at the end of the inquiry is the dualism we smuggled in at the start.
Subjectivity is a perspective, not a substance
Rovelli writes: “Subjectivity is not mysterious; it is just a special case of a perspective.” This is the right reframing. The phenomenal first-person is not a residue left over after the physical story is told; it is the same story told from inside the system telling it. RSP makes the same move under a different name. The strange loop at Level 3b is not a representation of experience; it is the experience, viewed from within the recursive loop. Two perspectives on the same events. Not two kinds of events.
The zombie argument is self-defeating
Rovelli’s handling of zombies is sharper than most of the analytic literature. A philosophical zombie, by stipulation, would claim to know what subjective experience is — otherwise it would be empirically distinguishable from a human. So during introspection, physical processes in the zombie’s brain would convince it of having consciousness. The same physical processes happen in me. If I trust my introspective conviction that I am conscious, knowing that the same conviction would be produced in a zombie that is not, then introspection cannot serve as evidence for non-physical experience. The argument bites before ontological commitments are settled. RSP arrives at the same verdict by a different route — a zombie with the full three-level recursive architecture is not coherently conceivable because phenomenology just is the architecture, and stipulating the architecture while subtracting experience is a category error like stipulating rotation without movement. Both arguments work. They work even better together.
Qualia are names for processes
Rovelli: “‘Red’ is the name of the process we generally undergo when we see or remember or think about the color red. We do not need to explain why it looks red for the same reason that we do not have to explain why the animal that we call ‘cat’ looks like a cat.” This is anti-qualia-realism done correctly. There is no meaning-independent raw quale of redness waiting to be explained over and above the integrated process. RSP makes the same commitment in different vocabulary: qualia are what recursive social self-modeling at Level 3b feels like from within, not a separately existing phenomenal substance the architecture happens to produce.
Rovelli and RSP both reject the framing of the hard problem itself. They both reject zombies, qualia-realism, and the view from nowhere. They both treat phenomenology as something a brain does from inside, not something it has alongside. The disagreement is not whether the hard problem is real. It is what dissolution requires.
The Hole in the Subtraction Move
The fault line runs through one of Rovelli’s most elegant sentences:
“These entities are not obtained by addition to a physical state, but by subtraction from a complete physical account.”
The move is beautiful. The mental is not something extra layered on the physical; it is the physical described at a higher level of grain, with detail dropped until only the salient features remain. Nothing supernatural. No new ontology. The soul is real because the high-level description is real, and the high-level description is a way of looking at the same matter, from within it.
The problem is that subtraction needs a criterion. What privileges the mental carving? Why is the high-level description that picks out recursive self-modeling the one that constitutes consciousness, rather than a thermodynamic description, or an ecological one, or a description at the level of immune response? Rovelli says the mental description captures “salient characteristics,” but salience is doing enormous work here without a principled account of what makes a characteristic salient for consciousness specifically.
Without that account, the subtraction view faces a dilemma. Either any sufficiently high-level description of a physical system constitutes a perspective — which slides toward a trivial panpsychism in which thermostats have inner lives — or there is a principled criterion for which high-level descriptions are the relevant ones, in which case Rovelli owes us that criterion. He does not provide it. He cannot, from where he is standing, because his argument is metaphysical-epistemological all the way down. It tells us the gap is malformed. It does not tell us which physical systems have perspectives, or why.
This is the place where RSP earns its keep.
What Architecture Adds
RSP is a specification of which high-level description is privileged and why. The privileged description is the one that tracks a particular organizational structure: core affect integrated with schema-organized prediction, accessible to a recursive self-model that predicts its own predictions. This is not arbitrary. It is grounded in three things subtraction alone cannot provide.
An evolutionary story: social-prediction machinery, evolved for modeling other minds, was recruited and turned inward to model the modeler. The architecture has a specific phylogenetic and ontogenetic order — Level 1 before Level 2 before Level 3 — that is not arbitrary but enforced by the cascade-control constraint that inner loops must stabilize before outer loops can be added. The Level 1→2→3 sequence is reflected in developmental milestones (interoceptive regulation in the first six months, schema formation by eighteen, explicit theory of mind by four or five) and recapitulated in evolutionary order across mammals, birds, and primates.
A mathematical convergence proof: the recursive loop at Level 3b stabilizes at a fixed point M* rather than generating infinite regress, because the self-modeling operator is a contraction mapping in the relevant function space. Without this, the strange loop would either diverge or collapse to triviality. With it, the architecture is shown to converge to a stable self-model under specified conditions, and the stability margins can be measured. The math is in Appendix D and Appendix E.
A pattern of clinical dissociations: thought insertion, depersonalization, schizophrenic loss of agency, frontal damage with preserved sensation but impaired self-modeling. These dissociations selectively disrupt exactly the architectural features RSP identifies as constitutive. Appendix G enumerates twenty-eight such mappings across eight clinical and empirical tables. The architecture predicts them; the subtraction view does not predict anything because it does not specify which features matter.
Without this specificity, Rovelli’s subtraction risks collapsing into eliminativism by another name. Qualia are “just names for processes” — but we have no account of why these processes rather than others, and the felt character of experience goes unasked rather than explained. The hard problem dissolves into nothing because nothing has been said. RSP says what has been dissolved into.
| Rovelli (relational physicalism) | RSP (architectural functionalism) | |
|---|---|---|
| What constitutes consciousness | A high-level perspectival description, obtained by subtraction from the complete physical account | A specific three-level recursive architecture: core affect, schema-organized prediction, recursive social self-modeling |
| Why this description rather than another | Salient characteristics — unspecified criterion | Evolutionary, mathematical, and clinical specificity of the recursive loop |
| What current LLMs lack | Unclear — perspectivalism applies to any system from within | Cascade feedback control, intrinsic motivation, recursive social self-prediction with the right contraction properties |
| Empirical predictions | None directly — the argument is metaphysical | Seven (Section IX), thirty clinical mappings (Appendix G), plus species-level predictions |
What Rovelli Gives RSP
The exchange is not one-directional. RSP’s constitutive-identity claim faces a version of the very question it aims to dissolve. Even granting the three-level architecture, a tenacious interlocutor can still ask: why should this architecture feel like anything? Why does recursive self-modeling at Level 3b generate phenomenal character rather than merely producing sophisticated self-reports in the dark? No amount of functional specification settles this question, because no functional vocabulary is itself phenomenal.
Rovelli short-circuits the question at the meta-level. Asking “why does this architecture feel like anything?” presupposes a view from nowhere from which we could observe the architecture running and then ask whether experience is additionally present. There is no such vantage. Any account is always from within the system. The self-model just is the perspective; it is not a representation of a separately existing phenomenal state that might or might not be present.
This is the move RSP needs to make its identity claim genuinely constitutive rather than merely stipulative. Without Rovelli’s perspectivalism, the constitutive identity at Level 3b looks like a relabeling: we describe the architecture, then we describe the phenomenology, then we assert without further argument that they refer to the same thing. With Rovelli’s perspectivalism, the identity is what the description looks like from inside the system being described. The relabeling is not a sleight of hand. It is the structure of any perspectival account.
RSP needs Rovelli to block the residual gap. Rovelli needs RSP to specify what closes it. The two arguments are not redundant. They are the two halves of a dissolution that neither half alone delivers.
Zombies, Qualia, Mary — Sharpened
On the three classical battlegrounds, the two views attack from different angles that together cover the field.
Zombies. Rovelli’s argument is epistemic and works from inside the scenario: introspective conviction cannot evidence non-physical experience because a zombie would have the same conviction. RSP’s argument is metaphysical and works from outside the scenario: a zombie with the full architecture is incoherent because phenomenology is constituted by the architecture. A Chalmersian can resist Rovelli by retreating to modal intuition independent of introspection. A Chalmersian can resist RSP by rejecting the constitutive thesis at the entry point. But the two arguments together leave no safe ground.
Qualia. Rovelli’s move is deflationary — “red” names a process, no further explanation needed. RSP’s move is constructive — qualia are what the recursive loop feels like from within, and the architecture specifies which loops have felt character and why. Deflation without construction risks looking like a question-begging relabeling. Construction without deflation risks a homunculus or an unexplained phenomenal residue. Together, the relabeling is non-arbitrary because RSP supplies the content of what is being relabeled, and the construction is non-question-begging because Rovelli establishes that the system does not need to produce experience as an additional output.
Mary. Neither paper addresses Jackson’s knowledge argument directly, but the joint position is clear: Mary learns no new fact about qualia because there are no qualia-facts independent of the perspectival-architectural process. What she gains is not new propositional knowledge but a new perspective — she now instantiates the recursive self-model that processes red, rather than merely modeling it from outside. Rovelli supplies the epistemological frame. RSP supplies the architectural content of what she now instantiates. The room dissolves the way the gap dissolves: by recognizing that the question presupposed a vantage that does not exist.
The Honest Disagreement
There is one tension that does not dissolve into complementarity. Rovelli’s perspectivalism is, in principle, sufficient for dissolution on its own terms. If the view-from-nowhere is incoherent, the hard problem dissolves without any architectural story. From this angle, RSP’s three-level cascade is interesting but unnecessary for the dissolution argument strictly construed. The architecture is good cognitive science. It is not philosophically required to make the gap go away.
The RSP reply is that Rovelli’s dissolution is too thin. It removes the bad question without answering the good one: which physical systems have perspectives, and why? A view that dissolves the hard problem without telling us whether a thermostat, a slime mold, an octopus, an infant, a sleepwalker, or GPT-7 has a perspective has not done the work. It has only shown that the work is permitted. RSP’s architecture is what does the work.
Both replies are correct. The disagreement is not about whether the hard problem is genuine — both say it is not — but about whether philosophical dissolution alone suffices, or whether it must be paired with a positive mechanistic account to be credible rather than merely asserted. My bet is the latter. Without the architecture, the dissolution is a posture. With it, the dissolution is a result.
Where This Leaves the Soul
Rovelli closes with a sentence I will steal. “The soul is not metaphysically different from the body. We are all parts of nature, like anything else in this sweet world.” This is the right way to land the argument. The soul is real. Anti-dualism does not eliminate it; it locates it. Where it is located, however, matters.
For Rovelli, the soul is a perspective — the brain experiencing itself from within, picked out by the right high-level description. For RSP, the soul is an architecture — the specific three-level recursive structure that constitutes the perspective Rovelli describes. These are not rival accounts. They are the same account viewed from two ends. Rovelli works top-down from physics-of-perspective. RSP works bottom-up from cognitive architecture. They meet at the point where the hard problem used to be.
Section X of “Beyond the Bat” organizes the field into twelve families (A–L). Rovelli does not fit any of them cleanly. The nearest neighbors are Family J (Recursive/Evolutionary Theories), where Trukovich’s R2R explicitly invokes Metzinger’s perspectivalness — a concept Rovelli pushes to its conclusion without taking on R2R’s architectural commitments — and Family H (Phenomenological Theories), the natural home for any perspectival argument about consciousness. The closest aligned theory, on the architectural side, remains Barrett’s constructed emotion (Family G). But Rovelli’s view is properly its own family: relational/perspectival physicalism — a category Section X does not currently name and should add. The recommended classification is closely aligned. RSP and Rovelli are not two theories of consciousness. They are one theory at two levels of description.
Rovelli is right that there is no hard problem. He is also right that the path to seeing this runs through giving up the view from nowhere. What he does not say — because his argument does not require him to — is that not every perspective is the kind of perspective that counts. Some physical systems instantiate the recursive loop that turns prediction back on the predictor. Most do not. The architecture is what tells you which.
There is no hard problem of consciousness. There is a hard architecture. Building it is the work that remains.