Study notes: Experience, Selfhood, and the Organization of a Mind

What organization makes a mental occurrence belong to a subject, and when does representing that belonging help constitute experience? The proposal: an occurrence belongs to the subject whose regulation partly constitutes it. Whether representing the belonging adds anything to experience then becomes a question for investigation, not definition. Yudkowsky proposes that subjective experience depends on background self-modeling. The author suspects that some of the capacities Yudkowsky emphasizes bear more on personhood than on phenomenal consciousness. He calls the suspicion a hypothesis to test.

The higher-order state is itself the phenomenally conscious state. Its descriptive content determines what is experienced. Its pointer content targets particular mental states without appearing in experience, and can keep activity available and route it. Brown even allows generative feedback. So Indexical Unity overstated the contrast between his targeting and describing and the authors’ maintaining and regulating. The disagreement is whether those activities are constitutive. Brown predicts that pointer content could be removed without changing what is phenomenologically manifest. The authors treat some actual regulatory coupling as constitutive. Brown’s challenge: if the first-order state already has everything that explains experienced redness, acquaintance only reveals an experience, and if not, what does acquaintance contribute? The essay accepts the burden of specifying the organization.

Yu and Carruthers hold that phenomenally conscious states are all and only the globally broadcast, nonconceptual, analog-magnitude representations. Sensory qualities are Fregean modes of presentation: a property presented as a point in a similarity space constituted by the perceiver’s discriminative capacities. The essay finds this close to availability through character and to the geometry of content-tracking structure. Divergences: discrimination is only one effect of an organization that can be finer than any discrimination it supports. Their phenomenal states are a subset of accessed ones, while here broadcast belongs to experience only where it closes a regulatory loop. And content-tracking structure can be discrete, as felt ownership of a limb suggests. On grain: Brown fixes character by higher-order description, Yu and Carruthers by discrimination, the essay by the organization of the vehicle, which can be finer than both.

Externalist attribution says B’s uptake makes the originating process an experience located wholly in A. If adding a receiver creates another experience at A, the individuation of A’s experiences depends on events A cannot register. If it does not, the externalist must explain why downstream organization decides whether A’s activity is experiential while its multiplicity leaves A’s experiential identity untouched. The principle: relations to inputs and to consumers are constitutive only where they causally affect the occurrence, in practice where they close a loop back onto it. Broadcast to consumers that reason, report or act is access and is not constitutive. Broadcast that gathers information from which corrections return to the vehicles is the outbound half of the regulatory loop and is constitutive. The one-way channel fails by construction. B’s minds can still have their own experiences of what A sends.

The aspects are tracking of particular ongoing representations, assessment of what they carry and how reliable they are, and intervention on their persistence, weighting or coordination. The added task is phenomenal binding: holding features carried by many vehicles together as one field of co-present items. That requires gathering information from across the system before computing corrections. Variants: local, where corrections are computed within a region, close to Lamme’s recurrent processing theory; global, where they need information only broadcast can supply, so experience requires a workspace whose broadcast returns; hybrid, where local regulation fixes the character of features and global regulation binds them. The author leans strongly against the local variant and mildly favors the global over the hybrid. He does not build on the leanings. Yudkowsky also sides against the local, calling experience “a big complicated process, not a small complicated process”. None of the variants makes experience coextensive with access.

Yudkowsky sorts properties into those that affect a system’s inputs and outputs and properties of the material layer that cannot. The essay accepts the argument against substrate views and panpsychism: nothing in its proposal is a material ingredient idle with respect to what a mind notices and says. The two classes are not exhaustive. A third lies between them: differences in causal pattern that do not depend on the substrate and need not show up in input and output. His example of “tiny trained squirrels running memos around your brain” holds fixed the pattern of cause and effect, not the input–output profile. The essay locates experience in the third class. Treating the classes as exhaustive would individuate experience by its role in producing reports, which is the functionalist premise in dispute. One reason to reject it: exact character is finer than anything cognition retains.

Laid out in time, every physical causal process is acyclic, brains included: a correction acts on the vehicle’s later state. So what distinguishes recurrent processing cannot be a cycle. It is whether a persistent carrier exists, something individuated by its causal structure whose later states receive corrections computed from assessing its earlier ones. An unrolled network with separate weights may encode content differently at every layer, with no shared format. A residual network or transformer reads from and adds into one shared stream, and later layers partly compensate when an attention layer is ablated. An unfolding that preserves the carrier preserves the pattern. The difference is invisible to behavior and visible to intervention. The pattern must be assessed at the grain at which the carrier is corrected: fine enough to individuate the regulated vehicles, coarse enough to abstract from where their components sit.

The earlier verdict was too quick, and the question is now held open. Within a pass a transformer has a persistent carrier in a shared format, corrected at every layer. The author also drops the requirement of coupling to the world, on the internalist reading he favors. Whether these brief occurrences are experiential is open. Across tokens the key–value cache is read and never revised, and the only path back runs through the sampled token: earlier vehicles are consulted, not regulated. Scaffolding regulates records. Recurrent-depth and recirculating models have longer-lived carriers and are stronger candidates, with a problem of individuating occurrences. The J-space locates availability, not experience. Against Chalmers, who reads depth as an alternative to recurrent feedback, the essay holds that a residual stream corrected through depth has the relevant structure. Self-reports do not decide. Interpretability evidence would: whether corrections land on the continuation of the assessed carrier and are gated by its content.

They agree that consciousness is made of parts and that experience is not a primitive quale that could be subtracted without changing behavior. Yudkowsky disclaims a requirement of effortful reflection. Still, each step of his reasoning concerns self-ascriptive report: “to say, ‘I feel this pain’, you must have a notion of ‘I’.” For a fuller model he points to Drescher, on whom a mental event is conscious by its “smart recording and playback” in an internal memory, a “Cartesian Camcorder”. For Drescher what makes an event experiential is being recorded and replayed for planning, which is access through memory. For the essay it is the regulation that constitutes the event, recorded or not. Brown’s three issues: which subject has the state, how the state is represented as belonging to that subject, and whether belonging figures in what the experience is like.

The achievements: an organization’s having mental states, its representing those states as its own, and its governing itself through that representation. A representation attributing activity to “me” does not settle belonging by its content. The criterion: an occurrence belongs to the subject whose regulation partly constitutes it. B is not the subject of A’s activity whatever B’s self-model says, because B regulates nothing in A. One occurrence constituted partly by two regulatory economies belongs to two subjects, which describes the sharing hypothesis. And the minimal self left when projection through time is removed from personhood is the regulatory economy of a present episode.

Perspectival subjectivity is the unity and centering of a binding economy: contents present together in one field organized around what the regulation holds in focus. It needs no represented viewer and no self-model. Phenomenal Consciousness as Mode of Being treated perspectivality as viewer-centered format, which was too deflationary, since format is a feature of content. Indexical Unity located perspective in self-directed targeting, which remains necessary and is not sufficient. Perspectival subjectivity thereby leaves the family of self-representation, where cognitive and agentic subjectivity remain, so the three are no longer parallel axes. For-me-ness is this perspectival subjectivity and belongs to every experience. Mineness is the layer agentic self-representation contributes, and it is contingent, as depersonalization shows. Yudkowsky’s formulations concern what a mind needs to say “I am sad”, which is me-ness and mineness. A self-modeling requirement may draw its plausibility from running these together. Whether binding into a centered structure occurs outside the experiential mode is left as an empirical question.

The two are made independent: both require mindedness and neither is constitutively defined through the other. The independence is constitutive, not statistical, since their mechanisms may overlap. Evidence for self-governance then counts as evidence for personhood without passing through a verdict on qualia. This retires the category of an aspiring person from the original personhood essay: a being with understanding and self-legislative agency that lacked homeostatically grounded knowledge. Dropping it gives up that form of epistemic grounding as a compulsory entrance requirement. A being capable of substantive self-governance can be a person without qualification, and a being can have experiences that matter without personal self-legislation. The essay calls this a counterproposal to Yudkowsky, who explicitly intends a theory of experience.

The earlier proposal grouped maintaining current activity with enforcing commitments across time. The criterion separates these as the organization of a present mental episode and the organization of a personal life. It also divides closure itself. The “this” form, demonstrative targeting of the system’s own states, comes with experiential regulation, which must track the vehicles it regulates. The open question is whether the “mine” form, representing those states as one’s own and using that in self-governance, is a natural achievement at all. Minimal selfhood might add only independently variable capacities, or identify a further organization, or gather steps on the way to personhood. What bears on it is dependence and intervention: whether the capacities sustain one another, and whether disrupting their coordination produces a change beyond the loss of one component. A minded system without a self is a test case, not an established counterexample.

• Toward Brown: phenomenal character reliably follows higher-order descriptive content despite interventions that disrupt the coupling the authors call constitutive.
• Toward a first-order account: the regulatory activity explains experience while representations of mental states explain only access and self-knowledge. If the regulating process does not represent its vehicles as representations, the account is first-order in Brown’s sense, and the authors should say so.
• Against carriers: interventions that dissolve a persistent carrier while preserving behavior change nothing an account of experience should track.

The commitment kept: experience must have a determinate enough concrete constitution for questions about its occurrence, extent and owner to be answerable. Whether the organization is best understood as acquaintance, a workspace whose broadcast returns, or a higher-order version of the proposal remains open.