David J. Chalmers (2018), Journal of Consciousness Studies 25(9–10), 6–61. Imprint Academic. The target article of a JCS symposium; the replies followed in a 2019 issue (an editorial introduction, “Debates on the Meta-Problem of Consciousness”, is at Ingenta, imp/jcs/2019/00000026), and the author’s reply, “How Can We Solve the Meta-Problem of Consciousness?”, appeared in JCS 27 (2020), pp. 201–226 (per his papers page). Neither was read.
Key takeaways
- The problem (intro, §1). Explain problem intuitions: dispositions to judge that there is a gap between physical processes and consciousness. They come in four central kinds: explanatory (gap, anti-functionalist), metaphysical (dualist, fundamentality), knowledge (Mary; other minds) and modal (zombies, inversion). Neighbouring kinds include distribution intuitions, such as “robots are not conscious”. It is an easy problem because problem reports are behaviour. The explanation must be topic-neutral, meaning it does not mention consciousness, though it may be representational, rational, historical or structural.
- Thirteen candidate solutions (§2). Introspective models; phenomenal concepts; independent roles; introspective opacity; immediate knowledge (with acquaintance, presentation and revelation); primitive quality attribution (Pereboom); primitive relation attribution (cf. Graziano’s attention schema); introjection; the user illusion (Dennett); the use–mention fallacy; underestimating the physical; historical and cultural explanations; others. A recurring test is “the problem of belief”: a proposal that would generate problem intuitions about beliefs too explains nothing distinctive.
- The proposed picture (end of §2). Introspective models with opaque concepts, largely independent of physical ones, attribute primitive mental relations to primitive perceived qualities, with a sense of immediate knowledge and acquaintance. The paper calls it neutral between realism and illusionism, and “closer to a fleshed-out explanandum … than an explanans”.
- The meta-problem challenge (§3). If mechanism M is the basis of consciousness, M should play a central role in producing judgments about consciousness. This tells against IIT: a zero-Φ simulation would make the same reports. It also tells against biological theories, whose silicon duplicates would report alike. It is an open question for global workspace and higher-order theories, a further aspect of the combination problem for panpsychism, and a failure for analytic functionalism.
- Six reactions (§4) and debunking (§5). Non-reductive: no solution, correlationism, realizationism. Reductive: strong illusionism, or consciousness as a lower-order or a higher-order meta-problem state. A debunking argument for illusionism and a coincidence argument are each set out, and three replies are given. Realizationism is favoured: on it, consciousness causally realises the meta-problem processes. But “it is hard to avoid a sense of coincidence entirely”.
- Illusionism (§§6–7). Weak illusionism does not dissolve the hard problem, since that problem needs only “the basic fact that there is something it is like to be us”. So illusionists who want to dissolve it must be strong illusionists. Strong illusionism is coherent, because an illusion can be a judgment and not an experience, but it is “empirically false”: people feel pain.
Standing in the record
Filed on 2026-10-02 at the owner’s request for David Chalmers in this
record. It is the hard problem’s sequel. It reframes the problem as a
tractable question about cognition, which is why epistemology and
cognition follow consciousness in its tags.
On the hard problem. Its introduction restates LIT-406’s analysis, and it turns it into a programme. It is the record’s fullest treatment of illusionism. Frankish’s illusionism, the target of §§6–7, is the stance of LIT-103, which charges Veit with leaving the hard problem “as hard as ever”. This paper’s §6 is the general version of that charge, aimed at weak illusionists. Hofstadter’s I Am a Strange Loop (LIT-423, unread) is cited in footnote 13 as “a sort of illusionism based in self-models and self-reference, primarily directed at illusions about the self”. Minsky’s Emotion Machine draft (LIT-432, read in NOTE-348) is not cited. By my reading it belongs with the deflationary proposals the paper surveys.
On AI consciousness (THEORY-023). Footnote 26 bears directly on it. A simple theorem-prover agent (Muehlhauser and Shlegeris) “mirrors at least some aspects of our phenomenal reports, while being so simple that most people would deny that it is conscious”. If that pattern continues, “we should probably be cautious” about tests that take puzzled phenomenal reports as evidence of machine consciousness. That is the mimicry worry in THEORY-023 (after Birch, LIT-111, and Schwitzgebel, LIT-191), applied to the one kind of behaviour often proposed as hardest to fake. The §3 challenge also applies to the indicator approach of LIT-056. An indicator taken from a theory whose basis M does not explain our reports inherits the challenge. That application is mine.
On the society-of-mind bridge (THEORY-043). Two bearings, both my inferences:
- Against IIT as the architectural criterion. The THEORY counts IIT’s
verdicts (exclusion; modular networks splitting into small complexes) as
one of the architectural routes. This paper’s §3 argues that Φ plays “no
essential role in explaining phenomenal intuitions”, since a zero-Φ
computational isomorph reports alike. That is a reason, independent of
the group case, to doubt that Φ is the basis of consciousness. It
supplies part of what the THEORY’s
promote_whenasks for: a ground for or against a criterion that does not come from its verdict on groups. - A meta-problem test for groups. By the §3 challenge, a criterion for phenomenal consciousness should go with the mechanisms that produce problem reports. Whether a nation produces problem reports of its own, over and above its citizens’ reports, is a question none of the readings asks. It might separate a society of mind, which on Minsky’s account produces its owner’s reports, from a nation, whose “reports” about consciousness are its members’. The paper does not discuss groups.
On the extended mind (LIT-097). No direct bearing. The paper’s problem intuitions include “self intuitions”, but it leaves them aside.
Kammerer is cited for an illusion meta-problem (the “resistance problem”) and as a possible illusionist (2016). Neither is the Kammerer paper held here (LIT-409, 2015, on the US). The paper’s own earlier solution to the meta-problem is in The Conscious Mind (LIT-441).
It carries no instruction for machine-learning practice. Its research programme includes computational models of phenomenal reports, and footnote 26 cautions against report-based tests for machine consciousness. Neither instructs ML practice, and it does not belong in the anthology.
The exchange it continues. Dennett’s “Facing Backwards” (LIT-397) says that once every function is subtracted, only the conviction that something remains is left. This paper takes that conviction as its explanandum.