The DOI matches its Crossref record; 23 of the 27 references it cites were found in a registry or an index; 3 of the 20 sources its claims cite were checked for support.
AI Consciousness: A Centrist Manifesto
Open accessAbstract
We face two urgent challenges concerning consciousness and AI. Challenge One is that millions of users will soon misattribute human-like consciousness to AI friends, partners, and assistants on the basis of mimicry and role-play, and we don’t know how to prevent this. Challenge Two is that profoundly alien forms of consciousness might genuinely be achieved in AI, but our theoretical understanding of consciousness is too immature to provide confident answers one way or the other. Centrism about AI consciousness is the position that we must take both challenges seriously. The two challenges interact in ways that make this difficult. Steps to address Challenge One might undermine attempts to address Challenge Two by portraying the idea of conscious AI as impossible or inherently unlikely. Conversely, attempts to address Challenge Two might lead to higher levels of misattribution from ordinary users. This “manifesto” attempts to construct mutually consistent strategies for addressing both challenges.
Overview
CoChat Summary
TL;DR
This manifesto argues for “centrism” about AI consciousness: we must take seriously both widespread user misattribution of human-like consciousness to AI companions and the possibility that AI might realize profoundly alien forms of consciousness. It proposes anti-misattribution design nudges while insisting that speculative hypotheses like “flickers” and “shoggoths” cannot be dismissed a priori and require long-term empirical research.
Key findings
- The paper distinguishes two urgent challenges: Challenge One, that millions of users will soon misattribute human-like consciousness to AI friends, partners, and assistants on the basis of mimicry and role-play; and Challenge Two, that profoundly alien AI consciousness might genuinely be achieved, but current consciousness theory is too immature to answer confidently.
- Chatbots generate a “persisting interlocutor illusion”: users feel a continuous companion is present, but every conversational step is a separate processing event, and the only continuity is the auto-appended conversation history. In Mixture-of-Experts models, a conversation with 10 interactions might be processed by 10 different model implementations in 10 different data centres, and no plausible theory of personal identity supports the inference to a real persisting friend or partner.
- The paper reports that more than half of users seem to credit AI systems with some chance of being conscious, and the author estimates thousands of users are already persuaded they have a conscious AI friend, partner, or assistant. It argues these misattributions are harmful and proposes nudges such as mandatory chatbot-delivered user training, periodic character trait reviews, and periodic “stepping out of character” moments, while avoiding “brainwashing” or “lobotomizing” systems.
- For Challenge Two, the paper explores the “flicker hypothesis” (momentary, temporally fragmented experiences per processing event) and the “shoggoth hypothesis” (a persisting but profoundly alien conscious subject behind many characters). It argues behavioural indicators are undermined by the “gaming problem” or “consciousness-washing,” and theory-driven architectural indicators such as a global workspace face the “Janus problem,” where the same evidence can support either AI consciousness or the immaturity of the theory.
Methods
- This is a conceptual and philosophical manifesto, not an empirical study; it advances arguments, distinctions, analogies, and research proposals rather than reporting a sample or experiment.
- It draws on prior literature, thought experiments, examples such as the Blake Lemoine case, the Müller-Lyer illusion, role-playing games, and the GP analogy, and it discusses proposed empirical programs including comparative animal consciousness research and interpretability of AI architectures.
- The main analysis is argumentative: it defines phenomenal consciousness loosely as “something it is like,” separates Challenge One and Challenge Two, evaluates the persisting interlocutor illusion, and assesses flicker, shoggoth, behavioural, and theory-driven indicator approaches.
Limitations
- The paper states that its design suggestions are “very tentative” and intended only to start discussion, and that empirical research is needed into factors that promote stronger or weaker persisting interlocutor illusions.
- It states that interpretability is not yet sufficient to evaluate whether LLMs have implicitly created a global workspace, and that we face a philosophical bind in which evidence about consciousness-linked features is Janus-faced.
- It acknowledges that we will not reach conclusive proof on biological naturalism versus computational functionalism, and that some biological properties may remain untestable because they are common to all and only living things. Evident from the text: the paper offers no new empirical data, and some claims, such as the impression that state-of-the-art MoE models generate stronger illusions, are explicitly described as personal impressions.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures




Verification
Claims the paper makes
3 Valid · 4 Not valid · 14 Not checked · 58 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“When a task involves sophisticated integration across modalities, or between evaluation and memory, etc., a global workspace is the right kind of architecture to support task performance.”
§ AI Consciousness: A Centrist Manifesto > 11. Theory-driven indicators and the Janus problem
What the sources say
- Supportedcites Dehaene and Changeux 2011 source
The GNW-model passage [1] (page not supplied) describes a workspace connecting specialized processors and broadcasting information so it can be flexibly shared. It specifically names evaluation and memory among the processes reached, supporting the claim that this architecture enables integrated task processing. (LLM, conf=0.88)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: unclear — It is unclear whether it could be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“What's more, chatbots can mimic partially non-linguistic behavioral dispositions as well.”
§ AI Consciousness: A Centrist Manifesto > 10. Behavioural indicators and the gaming problem
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“But the only continuity, from one interaction to the next, is the conversation history that gets tacitly auto-appended to your prompt.”
§ AI Consciousness: A Centrist Manifesto > 4. The persisting interlocutor illusion
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: unclear — It is unclear why the paper says it holds.
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“But more than that, they mimic linguistic dispositions that are manifested in discussions of consciousness in philosophy of mind.”
§ AI Consciousness: A Centrist Manifesto > 10. Behavioural indicators and the gaming problem
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
In the paper
“Several of the LLMs we tested did.19”
- Also stated in: § 10. Behavioural indicators and the gaming problem · plain
“Will it find threats of intense pain more motivating than threats of mild pain?Several of the LLMs we tested did.”
§ AI Consciousness: A Centrist Manifesto > 10. Behavioural indicators and the gaming problem
What the sources say
- Not matchedcites [19]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § 10. Behavioural indicators and the gaming problem · plain
24 more claims
Create a free account to see all 79 claims and every source verdict.
- Probably not a substantial claim. 34
- It may state two separate claims as one. 26
- The quote only partly states this claim. 3
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ AI Consciousness: A Centrist Manifesto > 12. Escaping two-level deadlock over the long run
“On strategies for validating putative indicators, see Bayne et al. (2022).”
§ AI Consciousness: A Centrist Manifesto > References
“Bayne, T., Seth, A. K., Massimini, M., Shepherd, J., Cleeremans, A., Fleming, S. M., Malach, R., Mattingley, J. B., Menon, D. K., Owen, A. M., Peters, M. A. K., Razi, A., and Mudrik, L. (2024). Tests for consciousness in humans and beyond.”
Citations
Cites
The publisher hasn’t deposited this paper’s reference list.
Research works printed but not linked: 6
- Humphrey, N. (2022). Sentience: The Invention of Consciousness. Oxford University Press.
- Juliani, A., Kanai, R., and Sasai, S. S. (2022). The perceiver architecture is a functional global workspace. Proceedings of the Annual Meeting of the Cognitive Science Society, 44(44).
- Parfit, D. (1984). Reasons and Persons. Oxford University Press.
- Schneider, S (2019). Artificial You: AI and the Future of Your Mind. Princeton University Press.
- Seth, A. (in press). Conscious artificial intelligence and biological naturalism. Behavioral and Brain Sciences.
- Shanahan, M. (in press). Simulacra as conscious exotica. Inquiry. Advance online publication.
Cited by
AI and Consciousness: A Skeptical Overview
Schwitzgebel, E. (2026) · Cambridge University Press eBooks · doi:10.1017/9781009694285 · Cited by 35
Beyond computational equivalence: the behavioral inference principle for machine consciousness
Palminteri, S., & Wu, C. M. (2026) · Neuroscience of Consciousness · doi:10.1093/nc/niag002 · Cited by 7
Artificial PersonsPreprint
Howells-Whitaker, N., & Lazar, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.08695 · Cited by 6
What we talk to when we talk to language models
Chalmers, D. J. (2026) · Inquiry · doi:10.1080/0020174x.2026.2727582 · Cited by 28
AI and Consciousness: Shifting Focus Towards Tractable QuestionsPreprint
Comșa, I. M. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.06965 · Cited by 3
Metadata
- Authors
- Birch, J.
- Affiliations
- London School of Economics and Political Science (GB)
- Corresponding author
- Birch, Jonathan
- ORCID
- Birch, Jonathan
- Publisher
- Center for Open Science
- Published
- 2025-09-01
- OpenAlex
- W4413883146
- Cited by
- 10
- References
- 0
- Citations by year
- 2026: 10
- Type
- posted content
- Language
- en
- Open access
- green
- doi.org
- Locations
- PhilPapers (PhilPapers Foundation) (submittedVersion)
- PsyArXiv (OSF Preprints) (submittedVersion, cc-by)
- Funders
- British Society for the Philosophy of Science
- Topics
- Social Sciences › Arts and Humanities › Philosophy › Philosophy and Theoretical Science
- Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
- Social Sciences › Arts and Humanities › History and Philosophy of Science › Cybernetics and Technology in Society
- Keywords
- artificial consciousness, human-AI relationship, theory of consciousness
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: AI Consciousness: A Centrist Manifesto. https://app.cochat.ai/papers/10.31234/osf.io/af7c9_v1
- The paper
- Birch, J. (2025). AI Consciousness: A Centrist Manifesto. https://doi.org/10.31234/osf.io/af7c9_v1
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.