David J. Chalmers (2026), Inquiry, 1–30 (published online 16 September 2026) — DOI 10.1080/0020174X.2026.2727582. A preprint circulated earlier on PhilArchive (CHAWWT-8), linked from the author’s papers page.
Standing in the record
Filed on 2026-10-02 at the owner’s request for the selfhood and personhood literature. It is the paper in which Chalmers turns from whether an LLM is conscious to what, and who, the thing we talk to is.
Unavailable because no lawful full text could be reached (see the history
note), so no NOTE is filed. What the record can say without it comes from the
abstract of the published version, which reads in full: “When we talk to
large language models, who or what is our interlocutor? First, I address
some issues about how best to characterize the interlocutor in terms of
mental states. Second, I discuss questions in the philosophy of computation
about what sort of AI system an LLM interlocutor might be: a model, an
instance, a thread, or a persona? Third, I analyze some issues about
personal identity in LLM interlocutors. Fourth, I draw some conclusions
about issues of AI welfare and moral status.”
The individuation tag is for the second question, which is about what
makes an LLM interlocutor one thing (model, instance, thread or persona)
before any question of selfhood. personhood is primary for the
third, with self for the persona question. Both are read off the abstract and should be checked by a reading.
Where it would sit in the record:
- Chalmers’s “Could a Large Language Model be Conscious?” (LIT-411). In that paper, as read in NOTE-345, Chalmers suggests in passing that one LLM “can support an ecosystem of multiple agents”, as a reply to the disunity objection. The abstract’s “a model, an instance, a thread, or a persona?” looks like the worked-out version of that remark. A reading should check whether the new paper changes the earlier one’s credences or treats unified agency, one of its six obstacles, differently once the bearer is fixed.
- THEORY-023. That account is about whether consciousness can be settled. This paper’s welfare and moral-status conclusions may not need consciousness at all, if they rest on mental states characterised some other way. A reading should say which.
- Goldstein and Lederman (LIT-206). Their candidates for the bearer of attitudes are the model and the instance, and they choose the instance. Chalmers adds the thread and the persona. The two Inquiry DOIs are adjacent (…2727582 and …2727584); whether the papers answer each other is for a reading to check.
- Shanahan. The persona is the simulacrum of the role-play paper (LIT-460), and Shanahan’s 2025 paper (LIT-466) lists nearly the same candidate sites of an LLM’s “I” (model, process, instance, conversation). A search-engine snippet of a post by Chalmers says a later version of the preprint added a section on interlocutors as “characters, personas, or simulacra”, discussing “role-playing vs realization” and the simulators framework. That was not checked against the text.
- The self literature. Whether its personal-identity section is psychological-continuity reasoning (as in Birch, LIT-111), narrative (as in Dennett, LIT-442), or about self-models (as in Minsky, LIT-432) cannot be told from the abstract. Hofstadter’s “Is there an ‘I’ in AI?” (LIT-416), also unread, asks a related question in its title. A reading should place it among the classical criteria (Locke, Parfit, Williams, Lewis on fission) being filed alongside, and among the modern ones.
What a reading should check: which candidate Chalmers takes the interlocutor to be, and on what grounds; whether a persona can be a person, and persist across models or threads; and what the welfare conclusions depend on.
Boundary. By its abstract it is philosophy of mind and ethics, with no
instruction for machine-learning practice; not held in the anthology; not
tagged anthology-candidate.