NOTE-347: Chalmers — Absent Qualia, Fading Qualia, Dancing Qualia

Contribution

The paper turns the standoff over absent and inverted qualia, intuition against intuition, into an argument. A functionalist cannot win by appealing to intuitions about the China brain. But anyone who denies the functionalist conclusion for an isomorph must accept one of two strange things: sudden disappearance of experience, or a dissociation between experience and judgment in a system working perfectly. The paper makes the thought experiment precise (the switch, the ten-percent step) and states what is and is not concluded: natural, not logical, necessity. It also lists the gaps it leaves open.

Key insight

Experience and the judgments a system makes about it are tied together tightly enough that you can notice a significant change in your experience when you attend to it. An organisation-preserving swap cannot change judgments, by construction. So it cannot change experience either, unless experience can float free of everything the subject could ever say or think about it. The argument turns a premise about self-knowledge into a constraint on the psychophysical laws.

Assumptions

• Functional organisation (§1) is a set of abstract components, a set of states for each, and dependency relations among the components, inputs and outputs. A physical system realises it when it can be divided into components whose causal dependencies mirror the specification.
• Grain (§1). Fine enough to fix behavioural capacities and dispositions. “In the brain, it is likely that the neural level suffices, although a coarser level might also work.” The neural level is used for illustration, and “the arguments generalize”. The generality is asserted.
• Corresponding states (§1). Isomorphs must be in corresponding states at the time (so a sleeping twin does not count).
• Empirical, not logical, possibility (§2). The question is what the laws of nature allow.
• Feasibility in principle (§3). A silicon chip can perform a neuron’s local input–output function, with transducers. If it cannot, a silicon isomorph is impossible and the principle is not engaged, so this is not a premise the conclusion needs.
• Judgment supervenes on organisation (§§3–4). “Unless one is a dualist of a very strong variety, beliefs must be reflected in the functioning of a system”. So an organisation-preserving change cannot add a belief such as “I can’t see anything” or “my qualia just flipped”.
• Reliable self-knowledge (§§3–4). A rational, unimpaired, attending subject is generally right about its own experience, and can notice a significant change in it.
• Continuity of fundamental laws (§3, footnote). Known laws make one magnitude depend continuously on another, and “there is no way to compound continuity into discontinuity”.
• Empirical premise (§5). Some systems, me or some biological systems, are in fact conscious.

Key results

All are thought experiments and plausibility arguments. The author says that “these arguments do not constitute conclusive proof” (§2).

• Fading qualia (§3). If Robot, a silicon isomorph of me, lacks experience, then in the replacement series either:

  • qualia suddenly disappear: one neuron, or a few molecules, switches off a whole field of experience. That needs a brute discontinuity at an arbitrary point; or
  • qualia fade. Joe, halfway along, says “bright red” and experiences “tepid pink”; says he hears loud noise and experiences “a distant rumble”. He is “utterly out of touch with his conscious experience, and is incapable of getting in touch”.

  Everyday fading, in drowsiness or down the evolutionary tree, comes with a matching change in function. Joe’s does not. Blindness denial is set aside, because its subjects are impaired and Joe is not. The sorites charge is answered: the argument accepts a continuum and rejects the intermediate cases for independent reasons.

• Extension to populations (§3). The same series works with neurons replaced by tiny homunculi, or external normal-sized ones with radio links, “ending up with a network of homunculi that is essentially equivalent to the population controlling a robot”. The homunculi can then “double up”, keeping written records, until one homunculus does all the work, as in the Chinese room. “If the causal organization of the original system is preserved, even if it is only among a system of marks on paper, then the same arguments suggest that the system will have experiences. (Of course, we should not expect the homunculus itself to have the experiences; it is merely acting as a sort of causal facilitator.)”

• Dancing qualia (§4). Fading does not touch inversion, since intermediate inversions preserve the structure of experience. Instead, among the 10%, 20%, … 90% systems, two neighbours must differ noticeably, because red cannot become blue in ten unnoticeable steps. Install the neighbour’s silicon circuit as a backup with a switch. Flipping it changes experience but not organisation, so “I am not suddenly disposed to say ‘Hmm! Something strange is going on!’”. The same construction works against absent qualia, from vivid to tepid. If that were possible, qualia might be dancing all the time, since low-level physiology changes constantly and only function is held fixed by evolution. “The only place to draw a principled line is at the functional level.”

• Loopholes conceded (§4).

  • Speed: a much faster or slower isomorph cannot be swapped in.
  • History: a physical twin with a different past cannot be reached by a switch, though a fading series can vary history.
  • Very mild inversions: changes smaller than one noticeable step. The step could be at most 20–25% before personal identity is in question, but if colour depends on a small area, one swap covers it.
  • Unattended qualia: these could invert unnoticed.

  Each is called coherent but unmotivated.

• Nonreductive functionalism (§5). Logical necessity cannot be reached. Fading and dancing qualia are intelligible. And even if they were impossible, the premise that some brains are conscious is empirical, so only the conditional could be necessary. So functional organisation “suffices for conscious experience with natural necessity”. That is “just as compatible with certain forms of property dualism … as with certain forms of physicalism”. Consequence: systems duplicating our organisation are conscious “even if they are made of silicon, constructed out of water-pipes, or instantiated in an entire population”, and biochemistry is “at best indirectly relevant”.

Claims

id	claim	strength	support
C1	Absent and inverted qualia are logically possible, but whether they are nomologically possible is the open question; intuition does not settle it	moderate	§2: the “10^{11} neurons are equally counterintuitive” parity argument
C2	If absent qualia are nomologically possible, fading qualia or suddenly disappearing qualia are possible	strong as a conditional, given the replacement series	§3: construction of the series; continuity of laws for the second disjunct
C3	Fading qualia are implausible: a rational, unimpaired system cannot be systematically wrong about its own experience	moderate	§3: appeal to “every case with which we are familiar”; the disanalogy with drowsiness, evolution and blindness denial; no room for new beliefs in an unchanged organisation
C4	If absent or inverted qualia are possible, dancing qualia are possible	strong as a conditional, modulo the loopholes	§4: the ten-percent pigeonhole step and the switch
C5	Dancing qualia are extremely implausible	moderate–strong	§4: the noticing principle; the skeptical consequence that our qualia might dance all the time
C6	Fine-grained functional isomorphs of a conscious system have qualitatively identical experiences (organizational invariance), with natural necessity	moderate as a plausibility argument	C2–C5; “not … conclusive proof” (§2)
C7	The argument extends to a population of homunculi and to a single record-keeping homunculus, and in that case the system, not the homunculus, has the experiences	asserted extension	§3, one paragraph; it relies on organisation being preserved at each step, which is stipulated
C8	The arguments cannot establish logical or metaphysical supervenience of experience on organisation	strong	§5: the conclusion’s necessity is capped by the empirical premise
C9	The principle does not say which organisations give rise to experience, or which experience	the author’s own statement	§5

Concepts

• Organizational invariance — experience is invariant across systems with the same fine-grained functional organisation (§1).
• Functional isomorph — a system sharing another’s organisation at a grain fine enough to fix behavioural dispositions, in corresponding states (§1).
• Absent / inverted qualia — an isomorph with no experience / different experience (§2).
• Fading qualia — experience that degrades along a replacement series while function stays fixed (§3).
• Suddenly disappearing qualia — experience that vanishes at one step (§3).
• Dancing qualia — experience that flips with an organisation-neutral switch, unnoticed (§4).
• Nonreductive functionalism — organisation determines experience with natural necessity, without reduction (§5).
• Causal facilitator — the role of a homunculus who implements an organisation without being its subject (§3).

Composition and nesting. This is the paper’s main bearing on the record’s bridge. The §3 extension describes a whole made of minded parts: people with radios, each standing in for neurons. On the paper’s argument that whole has the experiences of the person it replicates, and the homunculi keep their own. The paper never states an anti-nesting principle or argues against one. Its result is incompatible with any principle that forbids a conscious whole from having conscious parts as such, because the replacement series passes through wholes of that kind with no organisational change. The single-homunculus case separates two relations a part can have to a whole. A part can implement the whole’s organisation, like the record-keeping homunculus, whose own experiences are not the system’s. Or a part’s own capacities can be what the whole’s capacities consist in. The paper’s verdict covers the first relation only.

Connections

• Block (LIT-407, NOTE-366). The paper’s target, cited as “Block (1980)” in the text and “Block (1981)”, the Readings reprint, in the references; the original is 1978. Its gloss, that Block “argues that it is bizarre to suppose that this would somehow give rise to a group mind”, is not Block’s ground: Block’s doubt rests on dissimilarity and design, and survives replacing every homunculus with a gate (n. 19). The paper’s population series is Block’s n. 17 case, a billion people each simulating a neuron, the one where Block says intuition “seems to founder”. Block’s main China system is a machine-table realisation, and the paper reaches it only through the asserted generality of grain.
• Searle. The water-pipes, the Chinese room, and The Rediscovery of the Mind pp. 66–67 for a fading scenario in which beliefs stay true but impotent. The paper rejects that scenario: an unchanged organisation has “no room” for the new beliefs.
• Schwitzgebel (LIT-159, NOTE-131). Not cited (it is later). It supplies the premises the conscious-US argument needs against neurochauvinism and contiguism, and it refutes absolute anti-nesting for isomorphs. It does not supply the step from isomorphism to similarity of kind; see Bearing.
• Chalmers’s later objection in Schwitzgebel §6 (NOTE-131). That principle says a conscious system’s capacities must come largely from the relations among its subsystems, not from the subsystems’ own capacities. Its test case is the “two-seater homunculus”. The 1995 single-homunculus verdict is consistent with it, because there the homunculus is a facilitator executing recorded relations, not a source of the capacities. The later principle reads as the restriction this paper’s §3 already observes. That reading is mine; neither text links them.
• Kammerer (LIT-409, NOTE-362). His functional anti-nesting principle works with supervenience on functional organisation, and notes that this is “compatible with Chalmers’s naturalistic dualism”. It survives this paper. A homunculus in a neuron’s role does not consciously represent the whole, in a role only it could fill, since a chip could fill it.
• IIT (LIT-445, NOTE-365). IIT’s claim that a conscious system can have an unconscious functional equivalent (C7 there) is a denial of this paper’s principle. The two cannot both be true for fine-grained isomorphs.
• Seth (LIT-135). Seth doubts substrate flexibility. This paper is the standard argument for it, for isomorphs only.

Bearing on the record

• THEORY-043. The question to settle is whether organizational invariance, if true, licenses Schwitzgebel’s upward move. It does not. It licenses the replacement cases only:
  • Licensed: a fine-grained functional isomorph of a conscious being is conscious with the same experiences. That holds whatever realises it, including a population of minded people (§3). So it refutes absolute anti-nesting, that is, Putnam’s stipulation and IIT’s exclusion where it would deny the whole, for isomorphs.
  • Not licensed: any system whose organisation is merely similar in kind or degree to a conscious one. The United States is not an isomorph of a human or a rabbit, and no organisation-preserving series reaches it. Schwitzgebel’s step (NOTE-131, C1 and C7) needs a principle about which organisations suffice. The paper explicitly leaves that open (C9).
  • Consequence for the THEORY: if invariance holds, the anti-nesting route survives only in functional forms like Kammerer’s. The dispute between a society of mind and a nation then becomes wholly architectural, which is the THEORY’s second route. Invariance cannot settle it, because neither is an isomorph of the other or of us.
• THEORY-023. The paper is the canonical argument for the conditional “an exact functional isomorph of a conscious system is conscious”. No current AI system is such an isomorph, so the paper does not touch THEORY-023’s claim that current evidence cannot settle AI consciousness.
• LIT-097. No direct bearing.
• No ML instruction; not for the anthology.

Limitations

• It is a plausibility argument throughout, by the author’s own framing (the Schrödinger’s-cat comparison, §2).
• The noticing premise is supported by familiar cases. An opponent can hold that self-knowledge is reliable only for the biological systems it was observed in, which is the question at issue.
• The generality of grain is asserted. The series is built neuron by neuron, and the move to coarser realisations, including Block’s machine-table China system, is not shown.
• The population and single-homunculus extensions take one paragraph each, and they stipulate that organisation is preserved throughout.
• History and speed are open loopholes, conceded.
• It says nothing about which organisations are conscious. So it cannot adjudicate any case that is not an isomorph: animals, groups, AI systems.

Open questions

• Is there a principled extension from isomorphism to similarity, for example a continuity principle over organisations and not over replacements? That is what would carry the paper’s force to Schwitzgebel’s US. The paper does not attempt it.
• Does the dancing-qualia argument survive a grain change, from neurons to machine-table states? That would decide whether it answers Block’s main case or only n. 17.

Corrections

• None to a seeded skim (there was no seed).
• Venue. The brief gave “Imprint Academic”. The volume is co-published: Imprint Academic (Thorverton) with a Paderborn house, given as Schöningh by Chalmers and as mentis by Metzinger. See the LIT history.
• Against the paper. Block’s argument is cited as 1980/1981. The original is 1978 (LIT-407). It is also characterised as resting on the bizarreness of a group mind, which Block disowns (n. 19).