Jonathan Birch (2025), PhilArchive preprint (BIRACA-4); v4 acknowledges a special issue of Philosophy and the Mind Sciences “in which this article appears” — https://philpapers.org/rec/BIRACA-4
Key takeaways
- Two challenges — users misattributing human-like consciousness to chatbots, and the real but untestable-for-now possibility of profoundly alien AI consciousness — must be met together. Birch argues this means nudging users out of the “persisting interlocutor illusion” without denying that AI could be conscious, while building research programmes, via comparative animal-consciousness science, that could test both computational-functionalist and biological-naturalist claims.
Seeded from the abstract and a skim, before the reading in NOTE-183. What follows is what the work says about itself.
Birch identifies two urgent challenges. First, very many users will soon attribute human-like consciousness to AI companions and assistants on the strength of mimicry and role-play, and no one knows how to stop this. Second, AI might genuinely achieve profoundly alien forms of consciousness, and consciousness science is too immature to say either way. “Centrism” is the position that both must be taken seriously. They pull against each other: countering misattribution by calling conscious AI impossible undermines the second project, and taking AI consciousness seriously can feed misattribution. The manifesto tries to set out mutually consistent strategies for both.
Standing in the record
Filed from the 2026-09-26 sweep of the papers-feed tracker for philosophy: 20 seconds of active reading over 1 session. It was filed Deferred, unread. NOTE-183 is the close reading of 2026-09-26, and it placed the work: Active — worth reading as the clearest statement of the “centrist” position, and it coins four concepts the AI-consciousness literature now uses (persisting interlocutor illusion, gaming problem, Janus problem, two-level deadlock). It is a manifesto: its design nudges are untested proposals, and its way out of the deadlock is a research programme that has “barely begun” (§12).
When seeded, before the reading in NOTE-183, the entry gave a priority for a deeper reading: high — the paper is load-bearing for any AI-consciousness or AI-welfare note in the record. The skim captures the structure but not the detailed arguments of §§11–12.
What the seeding asked a deeper reading to check (NOTE-183 is that reading):
- It is a central current position paper on AI consciousness attribution. It pairs with LIT-135 (Seth, biological naturalism), LIT-207 (Roberts, fictionalism), LIT-212 (Šekrst) and LIT-206 (Goldstein & Lederman on “instance agents”), and it could anchor a nucleation cluster.
- The persisting-interlocutor argument rests on empirical claims about serving infrastructure (MoE routing, multi-datacentre serving) that the paper itself says are guarded secrets. A deeper reading should check how much the identity argument depends on them.
- Check the latest version (v6, July 2026) and the journal version for changes, especially to §§9–12.
Access when seeded: philpapers.org and philarchive.org returned 403. Read the full preprint PDF (v4, dated 20 May 2026, 32 pp.) from the Wayback Machine capture of 2026-05-24 of philpapers.org/archive/BIRACA-4.pdf: abstract and keywords, all section heads, §4, §7, §9, §10, the opening of §11, the opening of §12, and §13 with the acknowledgements. A later “v6, 31 July 2026” exists according to a search snippet but was not reached. published: is the date of Birch’s EA Forum post announcing the preprint (28 Aug 2025). The PhilArchive v1 upload date was not reachable and is unverified. The journal DOI and publication date are unverified.