NOTE-336: Chalmers — Facing Up to the Problem of Consciousness

Contribution

The paper names and separates the hard problem from the easy problems, and gives a reason why the separation is principled rather than a matter of current ignorance. The easy phenomena are defined by functions, and experience is not. It then refuses the two common responses, deflation and mysterianism. In their place it sketches what a nonreductive but naturalistic theory would look like: fundamental psychophysical laws, constrained by principles that can be argued for now.

Key insight

Ask, of any phenomenon, whether “you have explained the function but not the phenomenon” is a conceptual mistake. For the gene, life, learning and reportability it is. For experience it is not. Whatever answers the question “why is this processing accompanied by experience?” must therefore be something other than more mechanism, because any mechanism is more structure and dynamics, and “from dynamics, one only gets more dynamics” (§5).

Assumptions

• Functional definability of the easy phenomena (§3). “All it means for reportability to be instantiated in a system is that the system has the capacity for verbal reports of internal information.” The same is asserted for wakefulness, the gene and life. It is asserted as a conceptual fact, illustrated, not argued from a theory of concepts.
• Conceivability (§5). “Given any such process, it is conceptually coherent that it could be instantiated in the absence of experience.” So experience “is not entailed by the physical”. The step from conceptual coherence to failure of entailment is stated here and deferred to the book (LIT-441) for defence, including the reply to a posteriori necessity (Hill, Levine, Loar), which is a parenthetical “technical note” in §6.
• Causal closure (§6). “Physical laws already form a closed system”; psychophysical laws supplement rather than interfere.
• First-person data (§7). Regularities between processing and experience can be read off one’s own case and off verbal report, plus simplicity and homogeneity as non-empirical constraints.
• “Awareness” (§7). Information “directly available for global control”; to a first approximation, directly accessible and potentially reportable.

Key results

All are conceptual arguments, thought experiments or proposals. There are no data.

• Easy problems listed (§2). Discriminating, categorising and reacting to stimuli; integration of information; reportability; access to one’s own states; focus of attention; deliberate control; wake and sleep.
• Functional explanation (§3). Explaining a function is specifying a mechanism that performs it, and that is what neural and computational modelling do. “Function” means any causal role in producing behaviour, not a teleological one.
• Case studies (§4). Crick and Koch’s 35–75 Hz oscillations would explain binding and perhaps working memory, not why bound contents are experienced. A 1992 Discover interview is quoted in which Koch sets “subjective feelings” aside. Baars’s global workspace “shows promise as a theory of awareness, the functional correlate of conscious experience”, and leaves “why should global accessibility give rise to conscious experience?” unanswered. Edelman, Dennett (multiple drafts) and Jackendoff are given the same verdict in a sentence each. Five strategies: explain something else; deny the phenomenon (Allport, Dennett, Wilkes); claim to explain it (Flohr, Humphrey), where the step “ends up looking something like magic”; explain the structure of experience (Clark, Hardin); isolate the substrate.
• Extra ingredients (§5). Nonalgorithmic processing (Penrose), chaos, new neurophysiology and quantum mechanics (Hameroff) all explain more functions at best. The appeal of quantum theories is put down to a “Law of Minimization of Mystery”. Consciousness-collapse interpretations presuppose consciousness and do not explain it.
• Vitalism disanalogy (§5). Vitalists doubted that physical mechanisms could perform life’s functions, and accepted that explaining the functions would explain life. Here the functions are not in doubt, and the point is conceptual.
• Nonreductive explanation (§6). Take experience as fundamental, like charge after Maxwell. The theory is a set of psychophysical laws. The position is “naturalistic dualism”, “an innocent version of dualism”.
• Structural coherence (§7, principle 1). There is a direct correspondence between consciousness and awareness, and an isomorphism between their structures: colour space’s three dimensions and the visual system’s three axes; the visual field’s geometry; intensity. It does not capture intrinsic character, as the intelligibility of inverted spectra shows. It reinterprets work on neural correlates (Crick and Koch, Libet, Baars, Farah, Shallice) as hypotheses about the mechanism of awareness.
• Organizational invariance (§7, principle 2). “Any two systems with the same fine-grained functional organization will have qualitatively identical experiences.” Argued by reductio: two isomorphs with different experiences give a gradual neuron-to-silicon series in which two systems A and B, differing in under a tenth of the system, have different experiences. Install B’s circuit beside A’s with a switch, and flipping it changes experience with no possible functional trace, so “dancing qualia” go unnoticed. That is rejected as implausible. Absent and inverted qualia are “logically possible” but “empirically and nomologically impossible”. The principle is conditional: it says only that “if an isomorph is possible, then it will have the same sort of conscious experience”.
• Double-aspect theory of information (§7, principle 3). Information, in Shannon’s sense and Bateson’s “difference that makes a difference”, has a physical and a phenomenal aspect. Four supports are listed: changes in experience track informational changes; invariance needs an organisational property to attach to; it might explain structural coherence; and it fits the explanation of our judgments about experience. If all information has a phenomenal aspect, “perhaps a thermostat … might have maximally simple experience”. It is offered with a Russellian speculation that phenomenal properties are the intrinsic nature of the physical, “probably best ignored” for science.

Claims

id	claim	strength	support
C1	The problems of consciousness divide into easy (functional) problems and a hard problem of experience	moderate as a conceptual analysis	§§2–3: the functional-definability test, applied case by case
C2	No account of functions, whatever the mechanism, explains why the functions are accompanied by experience	moderate, given C4	§3, §5; it rests on the conceivability premise
C3	Current neuroscientific and cognitive theories (Crick and Koch, Baars, and others) address only easy problems	moderate for the two cases examined; asserted for the rest	§4: two worked cases, then one-sentence verdicts on Edelman, Dennett and Jackendoff
C4	Any physical process could conceptually occur without experience, so experience is not entailed by the physical	assertion here	§5; the defence, including the reply to a posteriori necessity, is deferred to the book
C5	A theory of consciousness should take experience as fundamental, with psychophysical laws that do not interfere with physics	proposal	§6, by analogy with Maxwell and charge
C6	The structure of experience is isomorphic to the structure of awareness (global availability)	moderate within familiar human cases	§7: colour space and visual-field examples; “not a logically necessary principle”
C7	Fine-grained functional isomorphs have qualitatively identical experiences	weak–moderate here	§7: a one-page sketch of the dancing-qualia argument; the full argument is in LIT-392
C8	Information has a phenomenal aspect, possibly everywhere it is physically embodied	speculation, so labelled	§7–§8: “more of an idea than a theory”

Concepts

• Easy problems / hard problem — problems of explaining functions / the problem of explaining experience (§2).
• Experience — what it is like to be an organism or in a state; the paper’s “consciousness” in the strict sense (§2).
• Awareness — the functional correlate: information directly available for global control (§§2, 7).
• Explanatory gap — Levine’s term, used for the gap between functions and experience (§3).
• Extra ingredient — whatever a theory adds to bridge the gap (§5).
• Naturalistic dualism — experience as fundamental, linked by natural laws, with physics closed (§6).
• Structural coherence, organizational invariance, double-aspect theory of information — the three principles of §7.
• Fine-grained functional organization — the abstract pattern of causal interaction between components (§7, defined more exactly in LIT-392).

Composition and nesting. The paper says nothing about wholes with minded parts. Two of its principles bear on nesting anyway. Invariance is indifferent to what realises the organisation, so it gives no reason to withhold experience from a system whose components are themselves minded. The double-aspect principle, unrestricted, puts experience wherever information is, at every scale. Neither is applied to groups here.

Connections

• Nagel (LIT-096) for “something it is like”, cited directly.
• Global workspace. Baars is the paper’s second case study. This puts the paper behind the record’s uses of workspace criteria: Butlin et al. (LIT-056), and Schwitzgebel’s reply to Chalmers (NOTE-131, §6), which invokes “fame” and workspace models of human consciousness. On this paper’s reading, a workspace criterion is a criterion of awareness and is connected to experience only through the coherence principle.
• List (LIT-401, NOTE-340) takes his awareness/experience distinction from Chalmers 1995. He treats the double-aspect view as the “3a” bridge principle and rejects it as too unrestricted.
• IIT (LIT-445, LIT-185). The paper lists “the integration of information” among the easy problems. That is a useful fact to set beside IIT’s identification of consciousness with integrated information. On this paper’s analysis, IIT’s Φ is either a measure of an easy-problem function or a proposed psychophysical law in the §6 sense. The pairing is mine.
• Seth (LIT-135, NOTE-169) uses the vitalism analogy that §5 argues against.
• Frankish (LIT-103) and the illusionist side are among the views §4 calls “deny the phenomenon”. The later meta-problem paper (LIT-425) takes them up seriously.

Bearing on the record

• THEORY-043. Two bearings. (1) Structural coherence would turn any grant of group awareness (List’s C1–C2) into a presumption of group experience, if it held beyond familiar cases. So the architectural route has to argue that coherence fails for groups, not just that groups lack a workspace. (2) The invariance principle, read here, is conditional on isomorphism. It does not by itself license the upward move; see the reading of LIT-392. Both are my inferences from the paper’s principles, which are not applied to groups.
• THEORY-023. The hard problem is the background reason behavioural evidence underdetermines experience. Organizational invariance is the conditional the functionalist side needs: an exact functional isomorph of a conscious system is conscious. It does not reach systems that are not isomorphs, which is every actual AI system.
• LIT-097. No direct bearing. It is consistent with the extended-mind paper’s exclusion of experience from its thesis.
• No ML instruction. It does not belong in the anthology.

Limitations

• The decisive premise, that any physical process is conceptually compatible with no experience, is asserted and deferred to the book.
• The case studies cover two theories in depth. The rest are dismissed in one sentence each.
• The invariance argument is compressed to a page; its loopholes (speed, history, mild inversions, unattended qualia) appear only in the full paper.
• Structural coherence is stated for “familiar cases”, and the paper does not say how far beyond them it is meant to reach.
• The double-aspect theory is labelled speculation by its author.

Open questions

• Does the coherence principle hold for systems unlike us: an animal, a machine, a group? If it does, the record’s group-awareness concessions have consequences for experience.
• What constraint, if any, picks out which information has a phenomenal aspect? The paper leaves this as “very much an open question”, and it is the question an anti-nesting principle would have to answer.

Corrections

• None to a seeded skim (there was no seed). The brief’s citation is correct.