David J. Chalmers (2019), in Matteo Colombo, Elizabeth Irvine and Mog Stapleton (eds.), Andy Clark and His Critics, Oxford University Press, ch. 1, pp. 9 ff. — DOI-10.1093/oso/9780190662813.003.0002
Standing in the record
Filed on 2026-10-02 at the owner’s request, as part of the extended-mind programme of David Chalmers. NOTE-356 is the close reading of the same day, and it placed the work: Active.
It carries LIT-097 further, and corrects it. The chapter opens by saying it will be “criticizing and revising the statement of the thesis in our original article”. It reconstructs the 1998 thesis (constitution by external entities “when those entities play the right role in driving cognitive processes”), concedes Farkas’s objection that this is near-obviously true of an implanted circuit and is accepted by Adams, Aizawa and Rupert, and replaces it with a perception-and-action version. It also withdraws LIT-097’s reply that Otto’s access to his notebook “is not perceptual at all”: that reply was “something of a stretch”, since Otto plainly sees the notebook. NOTE-180 recorded the same reply as the paper’s weakest step; this is its co-author agreeing.
It carries the foreword further. Chalmers’s foreword to Clark’s book (LIT-421) first set out the perception-action boundary as “by far the strongest objection” and first floated a bandwidth explanation of why consciousness does not extend. This chapter keeps the objection, makes it the definition of what is controversial, and drops the bandwidth explanation in favour of directness, conceding Karina Vold’s point that vision is a very high-bandwidth channel.
What it reaches and what it does not. On this statement the thesis covers standing beliefs and processes like mental rotation, linked by sensorimotor interaction. Wired-in circuits, brain-computer interfaces and intravenous stimulation fall outside it: the chapter places them with the subpersonal “extended circuit” cases, which everyone already accepts. That bears on Register’s account of BCIs as body parts (LIT-138): on this chapter’s terms a BCI is the uncontroversial kind of extension, not the extended mind. Consciousness is not reached by the perception-and-action route: the conclusion is that “even if there is extended cognition, there is no extended consciousness”. The chapter does allow that an extended circuit could carry consciousness, since directness is no obstacle there. So the denial is about notebooks, phones and screens, not about implants.
On the hard problem. The chapter stays neutral between physicalism and dualism by reading “partly constituted” as partial grounding, or, for a property dualist, as nomological dependence of consciousness on extended physical states. Its consciousness criterion, direct availability for global control, is the one Chalmers’s The Conscious Mind and his work on the neural correlate proposed as the physical correlate of consciousness; those are being filed in this record in the same batch. So the chapter uses a correlate, not an explanation of experience, and leaves the hard problem where it was.
On THEORY-043. That account says the upward move from minds to a
conscious group is refused either by anti-nesting or by an architectural
criterion, and that neither is yet principled. The directness criterion is
a candidate architectural criterion, stated by a different route, and it
has a feature that makes it bite on groups. Direct access, the chapter says,
“requires access that is not mediated by other mental states”. In a nation,
every link between parts runs through the citizens’ own perceptions and
actions, which are mental states of the parts. A society of mindless
agents in one skull has no such mediation. So, as I read it, the criterion
would separate Minsky’s case from Schwitzgebel’s, and it does so by
something close to an anti-nesting principle in functional dress. Two
cautions. Chalmers never applies it to groups. And its motivation has the
weakness THEORY-043 finds in the others: he offers it to explain an
independently held verdict (“there seem to be no plausible cases”), and
says it is not a knockdown. It is also not the objection Schwitzgebel
answers in LIT-159 §6 (NOTE-131): that was a relational-capacity
principle Chalmers put in correspondence. Applying the directness
criterion to both models is the kind of computation the theory’s
promote_when asks for, and nobody has done it.
On AI consciousness. Direct availability for global control is close kin to the global-workspace indicators in LIT-056 and so to the architectural half of THEORY-023’s dispute. For tools, the chapter’s verdict is that a user’s phone or assistant, reached by perception, can at most extend belief and cognition, never experience.
The extended-mind debate this chapter enters is the one held here in Adams and Aizawa (LIT-438), Rupert (LIT-422) and Clark’s reply (LIT-403); its §2 answers each of their objections in a sentence. It carries no instruction for machine-learning practice.