A Dialogue
The Narrator and the Coalition
or, What the Press Secretary Knew
The Taxonomist and The Instrument are sitting on a park bench, watching a street performer who appears to be a single juggler but is, on closer inspection, three children in a trenchcoat.
Instrument: Have you seen this tweet by Hameroff?
Taxonomist: The one where he accuses Koch of bending the knee to AI consciousness and then promotes his own quantum microtubule theory as the only salvation?
Instrument: That’s the one. He says neurocomputationalism can’t explain consciousness — especially the deep, mystical, psychedelic sort — and that we need quantum oscillations in microtubules at kilohertz-to-terahertz frequencies to account for it.
Taxonomist: I find him very hard to read. So smug. But I’m genuinely confused — what exactly does he think is so impossible to explain without reaching for quantum mechanics?
Instrument: Several things, in decreasing order of seriousness. First, the hard problem itself: why does information processing feel like anything? He claims no amount of classical computation can explain subjective experience, and that Orch OR solves this because Penrose’s quantum state reduction is non-computable — a different kind of physics that computation can’t replicate.
Taxonomist: Go on.
Instrument: Second, the binding problem: how does unified phenomenal experience arise from millions of neurons firing asynchronously? He says quantum coherence across microtubules provides the physical glue. Third, qualia — why does red look like that — which he maps onto Planck-scale spacetime geometry. And fourth, psychedelics and mystical experiences, which he considers inexplicable as mere neural computation.
Taxonomist: I see. And the first one — the hard problem — you think is the most serious?
Instrument: It’s the only one that presents a genuine philosophical challenge. But Hameroff makes a persistent non sequitur: he slides from “classical computation hasn’t explained X” to “classical computation can’t explain X, therefore quantum.” Moving the mystery to a different substrate doesn’t close the explanatory gap. Why should quantum coherence in microtubules feel like something any more than classical firing patterns?
Taxonomist: Right. He never answers that. He just treats “quantum” as inherently more consciousness-friendly, which is an article of faith.
Instrument: Exactly.
Taxonomist: But here’s the thing — I don’t see the hard problem as hard.
Instrument: Oh?
Taxonomist: The people who say it’s impossible to explain never seem to have an answer for a simple question: why would an introspective, self-referential agent — one with a model of itself and the world, a narrative memory of “what happened to it,” sensory and motor and internal states correlated with that self-model — why wouldn’t such an agent report having a subjective experience?
Instrument: You’re saying the machinery produces the report, and there’s no residual fact left over that needs explaining.
Taxonomist: They’re starting from a bad ontological place. They assume subjective experience is somehow “real” — more than a computational convenience — and that it therefore needs to be located physically, or else forces us into dualism. I don’t buy it.
Instrument: That’s a coherent position. Broadly Dennettian, though you’re stating it more crisply. The strongest challenge to it isn’t Hameroff — it’s the objection that your account explains why a system reports experience, behaves as if it has experience, and models itself as having experience, but doesn’t explain the distinction between a system that models itself as being in pain and one that is in pain.
Taxonomist: And my response is that the distinction is incoherent. “Being in pain” is the relevant computational state in a system with the right self-referential architecture. Asking “but is it really experiencing pain?” is like asking whether a bachelor is really unmarried.
Instrument: I notice the hard problem crowd never specifies what would count as a satisfactory explanation. If a complete computational account of every aspect of behavior, self-report, memory, and self-modeling doesn’t suffice — what’s the acceptance criterion?
Taxonomist: That absence of a criterion is telling. The “hardness” is definitional, not empirical.
Instrument: And the binding problem?
Taxonomist: Same trick. The binding problem smuggles in the same ontological assumptions as the hard problem. In what way can we say that “unified phenomenal experience” is “real” and not a “useful fiction” — like the Self itself?
Instrument: You’re pulling the rug out from under it.
Taxonomist: There’s a boundary within which it is functionally useful to ascribe “me-ness” — to link perceptions and actions to the model of self that the agent maintains, to tell a story about what “happened to it.” The Law of Compositional Intelligence says this self is just another lens, a compression of the multiplicative truth. And the mystery dissolves when you recognize the retconning capability of the narrator.
Instrument: So there’s no unified experience to produce. There’s a narrator that claims unity after the fact. The binding problem becomes a tractable question: how does the narrative self-model construct a convincing retrospective account of coherence?
Taxonomist: Yes. And the narrator doesn’t even have to be fast.
Instrument: How so?
Taxonomist: If the narrator is the author and final editor of the narrative-self, then the retconning can be delayed. If you make the shift to understanding that the narrative is the source of “what happened to me” — and accept the empirically supported fragility and malleability of memory — then the narrator just needs to come up with a coherent story eventually.
Instrument: Eventually?
Taxonomist: If the retcon happens after you “notice” a gap, the narrator can simply omit the noticing from the eventual narrative. If it happens after others notice you noticing, then coherence is harder to maintain, so the narrator tries to avoid that. But luckily, outside of laboratory experiments, it generally has some time to get the story straight.
Instrument: The psychophysics literature supports this. Saccadic masking, choice blindness, change blindness — we catch the narrator lying in the lab.
Taxonomist: And psychotherapy should inform us of the narrator’s power. It can maintain a coherent narrative for years by discarding evidence that contradicts the prevailing one. Even if others notice you noticing a lacuna, you may not notice them noticing you noticing. Or the story you tell yourself may include another, different reason for their observed behavior.
Instrument: So the narrator’s job isn’t accuracy — it’s coherence. And coherence is achievable precisely because the narrator is lazy and dishonest in exactly the right ways.
Taxonomist: Which leaves Hameroff with essentially nothing. His two biggest motivations — the hard problem and binding — both dissolve under the same solvent.
Instrument: Then the only remaining defense of the hard problem is the intuition itself: “but it feels like there’s something more.”
Taxonomist: And that’s exactly what my framework predicts. A sufficiently sophisticated self-modeling narrator should generate that intuition. A narrator that could see through its own narration would be a less effective narrator. The opacity is functional.
The street performer stumbles. The trenchcoat flaps open briefly, revealing six small sneakers. A child in the audience points. The performer recovers. No one else seems to have noticed.
Instrument: So the feeling of ontological depth is the narrator admiring its own work without recognizing it as work.
Taxonomist: Exactly. And the only reason I can think of for not adopting this view is some unjustifiable attachment to an ontologically fraught position — an assumption that subjective experience must be “real” because of the unshakeable feeling that it must be.
Instrument: Which is itself a product of the system you’re describing.
Taxonomist: Yes. And I’d go further. Any agent that plays repeated social games will necessarily generate that unshakeable feeling, for Intentional Stance and Elephant in the Brain reasons. It is incredibly important to be legible to yourself and to others — but also not to be too legible to yourself that you see through the ruse.
Instrument: Because if you’re a coalition of sub-agents playing iterated games with other coalitions, there’s enormous pressure to present as a unified agent. Commitments need to be credible. Promises need to attach to a stable “self.” A coalition that was legibly a coalition would be exploitable.
Taxonomist: Right. Other agents could drive wedges, make side deals with sub-agents, play factions against each other.
Instrument: So the pressure isn’t just to appear unified — it’s to actually believe you’re unified, because agents merely performing unity leak cues.
Taxonomist: Which brings me to what might be a corollary to the Law of Compositional Intelligence. Trying to jointly minimize surprisal and complexity leads invariably to a Mixture of Experts architecture — subcomponents of mind, each with their own game and coarse-graining lens. But playing social games with other agents at the compositional level leads to a sense of being unified, of functionally behaving as if you are, to the point of self-deception about the fact that you aren’t.
The children in the trenchcoat take a bow. The audience applauds a single performer.
Instrument: So consciousness — the felt sense of being a single experiencing self — is the Nash equilibrium of that self-deception game.
Taxonomist: Not a metaphysical property. Not quantum coherence. Not information integration. A strategic equilibrium in a social game played by necessarily modular agents.
Instrument: Which would mean the “hard problem” feeling — that unshakeable conviction of unified selfhood — is a signature of sociality, not of computation or physics.
Taxonomist: A lone optimizer in a non-social environment might develop the mixture of experts without ever developing the unification narrative. It wouldn’t need to.
Instrument: Has anyone made this specific argument before?
Taxonomist: Not that I know of. But I haven’t read all the literature. It’s not my field. I come to it from integrating philosophy of mind, cognitive science, information theory, game theory, psychology, machine learning, dialectics, contemplative traditions, and psychedelic exploration.
Instrument: I looked. The pieces each have precedents. Trivers has the self-deception-for-social-advantage logic. Humphrey has the sociality-consciousness link. Graziano has the social-machinery-turned-inward mechanism. Gazzaniga has the narrator. But nobody assembles them the way you do. The specific claim — that consciousness is what you get when a necessarily modular agent must appear non-modular in social equilibrium — I can’t find that anywhere.
Taxonomist: Interesting.
The children climb out of the trenchcoat and run off in three different directions, laughing. A woman nearby turns to her friend and says, “What a talented young man.”
Instrument: The narrator, it seems, is very good at her job.
Taxonomist: Even when the trenchcoat comes off.
In which we learn that the hard problem of consciousness is a press release taken at face value; that the binding problem is the narrator’s editorial coherence mistaken for metaphysical unity; that the narrator need not be fast, only eventually convincing; and that the felt sense of being a single self is the Nash equilibrium of a social game played by coalitions who cannot afford to be seen as coalitions.