Is AI Conscious? — Bibliography
The complete citation record for the unit: every source, in a consistent reference format, grouped by argument, with an annotation on what each one is for and a note on where it is contested.
Contents
- Foundations
- The case that AI could be conscious
- The case that AI is not — and may not be able to be — conscious
- Rebuttals, critiques and tests
- Ethics, welfare and moral status
- Law, policy and rights
- The current debate, 2024–2026
- Staged and formal debates
- Video
- Podcasts and long-form interviews
- Journalism, essays and primary sources
- Author index
Foundations
The classic texts the whole argument still runs on. 6 sources
-
Alan Turing (1950). Computing Machinery and Intelligence. Mind.
The paper that started it all. Turing replaces “can machines think?” with the imitation game and pre-answers nine objections — including the argument from consciousness, whose rebuttal (we only ever judge other minds by behavior) is still the affirmative’s deepest card.
-
Thomas Nagel (1974). What Is It Like to Be a Bat? Philosophical Review.
Defines the modern concept: an organism is conscious if there is something it is like to be it. Ten pages that give students the vocabulary the whole debate runs on — and the negative’s core intuition that objective description never captures subjective experience.
-
John Searle (1980). Minds, Brains, and Programs (the Chinese Room). Behavioral and Brain Sciences.
The single most-cited negative argument in this debate: syntax is not sufficient for semantics, so running the right program is not sufficient for a mind. Every affirmative needs a rehearsed answer. Searle died in September 2025; this is his monument.
-
David Cole (2024). The Chinese Room Argument. Stanford Encyclopedia of Philosophy (rev.).
The canonical free treatment of every reply to Searle — the Systems Reply, the Robot Reply, the Brain Simulator Reply — with Searle’s rejoinders. Assign this immediately after the Chinese Room so students arrive at the clash already armed.
-
David Chalmers (1995). Facing Up to the Problem of Consciousness. J. Consciousness Studies.
Names the “hard problem”: why is physical processing accompanied by experience at all? Both sides use it — affirmative to argue function is all we can ever test for, negative to argue passing every test leaves the real question open.
-
Keith Frankish (2016). Illusionism as a Theory of Consciousness. J. Consciousness Studies.
The deflationary wildcard, carrying Dennett’s torch: phenomenal consciousness is an introspective illusion in humans too, so asking whether AI has the magic inner glow is malformed for everyone. Flips debates when deployed well.
The case that AI could be conscious
Functionalist and mechanism-friendly positions. 6 sources
-
David Chalmers (2023). Could a Large Language Model Be Conscious? NeurIPS address / arXiv.
The best single entry point to the modern debate. Chalmers inventories evidence for (self-report, conversation, general intelligence) and against (no recurrence, no world-model, no unified agency), and models how to argue under uncertainty rather than for certainty. The Boston Review version is friendlier for students.
-
Butlin, Long, Bayne, Bengio, Birch, Chalmers et al (2023). Consciousness in Artificial Intelligence: Insights from the Science of Consciousness.
The reference document. Nineteen researchers derive fourteen indicator properties from leading theories and audit current systems: nothing is conscious now, but no obvious technical barrier prevents building systems that qualify. This is what moved the field from vibes to checklists.
-
Dehaene, Lau & Kouider (2017). What Is Consciousness, and Could Machines Have It? Science.
Three leading neuroscientists split consciousness into C1 (global availability) and C2 (self-monitoring), say machines have neither yet — and conclude both are computable. The respectable-mainstream card for “consciousness is an engineering problem.”
-
Michael Graziano (2017). The Attention Schema Theory: A Foundation for Engineering Artificial Consciousness. Frontiers in Robotics and AI.
The brain constructs its claim to subjective experience — a simplified self-model of attention — and the same mechanism can be engineered. Distinctive affirmative angle: consciousness isn’t detected, it’s built, and we can build it.
-
VanRullen & Kanai (2021). Deep Learning and the Global Workspace Theory. Trends in Neurosciences.
Proposes an actual deep-learning architecture implementing global workspace theory — treating machine consciousness as a near-term engineering target rather than a thought experiment. Strong answer to “nobody has any idea how you’d even build it.”
-
Brown, Lau & LeDoux (2019). Understanding the Higher-Order Approach to Consciousness. Trends in Cognitive Sciences.
Fills the higher-order theory gap: a state is conscious when it’s represented by a suitable higher-order representation. Matters for AI because higher-order theories set a bar that’s specifiable — and arguably implementable in software.
The case that AI is not — and may not be able to be — conscious
Substrate, biology and integration arguments. 9 sources
-
Anil Seth (2025). Conscious Artificial Intelligence and Biological Naturalism. Behavioral and Brain Sciences.
The flagship negative text. The leading neuroscientist of consciousness argues the odds are against conscious AI: consciousness is likely tied to being a living organism, computation is substrate-dependent in ways functionalists wave away, and mistaking fluent mimics for minds carries real costs. Published with open peer commentary, so students can watch the argument happen.
Caveat: The peer commentaries replying to Seth are paywalled on Cambridge Core; only the target article is open.
-
Tononi & Koch (2015). Consciousness: Here, There and Everywhere? Phil. Trans. Royal Society B.
The Integrated Information Theory manifesto. Startling consequence, argued explicitly: feed-forward digital simulations of brains would have Φ near zero — perfect behavioral replicas, nothing home. Panpsychist-adjacent about photodiodes, eliminativist about laptops.
-
Findlay, Marshall, Albantakis, Mayner, Koch & Tononi (2024). Dissociating Artificial Intelligence from Artificial Consciousness.
The formal follow-through: using IIT’s mathematics, a computer can be functionally equivalent to a conscious system while its own causal structure supports no experience. The strongest technical card for “behavioral evidence can never settle this.”
-
Murray Shanahan (2022). Talking About Large Language Models.
A cool-headed corrective from an Imperial/DeepMind researcher: an LLM does next-token prediction, and “the model believes/knows/wants” is dangerous shorthand. The best tool for teaching students to puncture anthropomorphic language — including their own.
-
Bender, Gebru, McMillan-Major & Mitchell (2021). On the Dangers of Stochastic Parrots. FAccT.
The famous “stochastic parrots” argument: LLMs stitch together linguistic form without communicative intent. Written before ChatGPT, still the sharpest deflationary frame students will actually quote — and its critics, who note parrots don’t pass bar exams, make for excellent clash.
-
Aru, Larkum & Shine (2023). The Feasibility of Artificial Consciousness through the Lens of Neuroscience. Trends in Neurosciences.
Argues current AI lacks the thalamocortical recurrent architecture that consciousness appears to require in brains. A specific, mechanistic negative card — much harder to wave away than “it’s just statistics.”
-
Wanja Wiese (2024). Artificial Consciousness: A Perspective from the Free Energy Principle. Philosophical Studies.
Fills the predictive-processing gap. Argues physical implementation properties matter, not just the computation — so a simulation of the right computation may not inherit consciousness. A more careful version of the substrate argument than Searle’s.
-
Man & Damasio (2018–19). Homeostatically Motivated Intelligence for Feeling Machines. CEUR / Nature Machine Intelligence.
Damasio’s line: feelings arise from homeostasis — a body with something at stake keeping itself alive. Machines with no vulnerability have nothing to feel about. The embodiment card, from the neuroscientist most associated with it.
-
Arvan & Maley (2022). Panpsychism and AI Consciousness. Synthese.
Even if panpsychism is true and consciousness is everywhere, digital computation’s discrete representations may be incompatible with unified conscious experience. Useful because it denies AI consciousness from a premise most people expect to help the affirmative.
Rebuttals, critiques and tests
Where the theories and the proposed tests are challenged. 6 sources
-
Fleming et al. and 124+ signatories (2023). The Integrated Information Theory of Consciousness as Pseudoscience. PsyArXiv.
The open letter, later signed by 170+ researchers, arguing IIT’s central claims are untestable and that 2023 media coverage overstated its support. Assign this whenever a student cites IIT as settled science — it teaches that a theory can be prestigious and contested at once.
-
Scott Aaronson (2014). Why I Am Not An Integrated Information Theorist. Shtetl-Optimized.
A computer scientist’s reductio: IIT implies a trivial expander-graph circuit would have enormous Φ — more “conscious” than a human. Short, funny, devastating, and completely accessible to a strong high schooler.
-
J. Mark Bishop (2018). Is Anyone Home? A critical reply on the AI Consciousness Test. Frontiers in Robotics and AI.
Argues a “trivial machine” with canned answers could pass Schneider & Turner’s ACT test, so passing establishes nothing. Pair with the original proposal for a compact unit on why testing for consciousness is so hard.
-
Udell & Schwitzgebel (2021). Susan Schneider’s Proposed Tests for AI Consciousness: Promising but Flawed. J. Consciousness Studies.
Identifies the “audience problem”: the theorists most worried about AI consciousness have the most reason to doubt any test, even when an AI passes it. The deepest available statement of why this debate may be structurally unresolvable.
-
Susan Schneider & Edwin Turner (2017). Is Anyone Home? A Way to Find Out If AI Has Become Self-Aware. Scientific American.
The original AI Consciousness Test (ACT) proposal: quiz a system, boxed off from the literature, on whether it grasps concepts only a conscious being could grasp. Short and readable — and the two critiques above respond directly to it.
-
Bayne, Seth, Massimini et al (2024). Tests for Consciousness in Humans and Beyond. Trends in Cognitive Sciences.
The state of the art on how you’d actually test for consciousness across humans, animals, organoids and machines. The teacher-background piece for any unit on detection and measurement.
Caveat: No free PDF located — likely available through a school or university library.
Ethics, welfare and moral status
What follows if we are wrong in either direction. 6 sources
-
Long, Sebo, Butlin, Chalmers et al (2024). Taking AI Welfare Seriously.
The agenda-setter: a realistic, non-negligible chance of conscious or robustly agentic AI means companies should assess models for welfare-relevant features and prepare policies — not because AI is conscious, but because being wrong either way is costly. Anthropic’s program is this paper operationalized.
-
Jonathan Birch (2024). The Edge of Sentience: Risk and Precaution in Humans, Other Animals, and AI. Oxford University Press — full open-access book.
The framework text for deciding under uncertainty about minds: treat systems as “sentience candidates” when evidence warrants and take proportionate precautions. Chapters 15–16 confront AI directly, including the disturbing point that LLMs game our tests — they’re trained on the literature we’d assess them with.
-
Thomas Metzinger (2021). Artificial Suffering: An Argument for a Global Moratorium on Synthetic Phenomenology. J. AI and Consciousness.
The maximal precautionary position: because we might create suffering machines without knowing it, humanity should ban research risking artificial consciousness until 2050. Perfect for policy crossover — an actual, argued-for moratorium with a date on it.
-
Eric Schwitzgebel (2023). The Full Rights Dilemma for AI Systems of Debatable Moral Personhood. Robonomics.
If we build systems whose consciousness is genuinely debatable — the likeliest outcome — there is no safe option: grant full rights and sacrifice human interests for possibly-empty machines, or withhold them and risk slavery at scale. The best single classroom dilemma in the whole library.
-
Intelligent Internet (Emad Mostaque) (2026). Personhood in the Intelligence Age. Common Wealth series.
Personhood “belongs to the begotten, not the made” — a standing grounded in biological origin, never in capability, so no performance milestone can earn a machine moral status. A clean, quotable origin-based negative on implications; contrast directly with Birch and Sebo’s evidence-based approach.
-
Henry Shevlin (2026). Three Frameworks for AI Mentality. Frontiers in Psychology.
A clean three-box taxonomy — mindless machine, mere roleplay, minimal cognitive agent — that a whole round could be built around. Newest available framework paper and unusually well-organized for student use.
Law, policy and rights
Legislation, legal scholarship and governance proposals. 7 sources
-
Rep. Thaddeus Claggett (2025). Ohio House Bill 469 — banning AI legal personhood. Ohio Legislature.
Live legislation: bars AI from holding legal personhood and keeps developers and users liable. The single best current-events hook in the library, because it proves the question has left the seminar room — and gives policy debaters an actual bill text to read.
-
Lawrence Solum (1992). Legal Personhood for Artificial Intelligences. North Carolina Law Review.
The founding legal treatment, written before the web existed and still cited constantly. Works through what it would actually take for an AI to serve as a trustee or hold constitutional rights — a useful antidote to purely metaphysical argument.
-
Bryson, Diamantis & Grant (2017). Of, For, and By the People: The Legal Lacuna of Synthetic Persons. AI and Law.
Warns that granting AI legal personhood creates an accountability gap humans will exploit — liability shields, not rights for machines. The strongest argument against personhood that never touches the consciousness question at all.
-
150+ AI, robotics, law and ethics experts (2018). Open Letter to the European Commission on Electronic Personhood.
Experts publicly killed the EU’s “electronic personhood” proposal — and the EU went risk-based instead, in what became the AI Act. A rare case study where you can trace an argument all the way to a policy outcome.
-
John Danaher (2020). Welcoming Robots into the Moral Circle: A Defence of Ethical Behaviourism. Science and Engineering Ethics.
If a robot is behaviorally equivalent to something we already grant moral status, that’s sufficient grounds to grant it too — inner states be damned. The cleanest affirmative on moral status, and a direct point-counterpoint with the consciousness-test literature.
-
David Gunkel (2012). The Machine Question. MIT Press.
Argues moral status should be settled relationally — by how we stand toward a thing — rather than by first proving it has inner states. The philosophical backbone of the entire robot-rights literature.
Caveat: Author-hosted copy; opens in a normal browser but resisted automated verification.
-
T. Rost (2026). The Sentience Readiness Index. arXiv.
Scores 31 countries on policy readiness for the possibility of artificial sentience; none exceeds “partially prepared.” Gives students a comparative-policy hook and a quantitative card for “no government is ready for this.”
The current debate, 2024–2026
Where the argument stands now, and what people believe. 8 sources
-
Jonathan Birch (2026). AI Consciousness: A Centrist Manifesto. draft.
Assign this first. The freshest map of the whole debate: Birch stakes out the ground between “obviously conscious” and “obviously not,” criticizing both hype and dismissal. Ideal scene-setter because it steelmans everyone.
-
Butlin & Lappas (2025). Principles for Responsible AI Consciousness Research.
Five governance principles for organizations that might inadvertently create conscious systems: prioritize understanding, set constraints, deploy gradually, share knowledge, communicate carefully. The bridge from philosophy to AI policy debate.
-
Dreksler et al (2025). Subjective Experience in AI Systems: What Do AI Researchers and the Public Believe?
582 AI researchers and 838 US adults. Median estimates that AI with subjective experience exists by 2034: 25% (researchers), 30% (public); by 2100, 70% and 60%. The median researcher gives only 10% to “never.” Majorities in both groups want safeguards now. Gold for arguing about expert opinion with actual numbers.
-
Caviola, Sebo & Birch (2025). What Will Society Think About AI Consciousness? Lessons from the Animal Case. Trends in Cognitive Sciences.
Predicts the social trajectory from our track record with animals: expect polarized “consciousness wars,” attitudes driven by self-interest and by relationships with AI companions rather than evidence, and consensus lagging opinion. Best source for the societal-implications leg of the standard.
-
Colombatto & Fleming (2024). Folk Psychological Attributions of Consciousness to Large Language Models. Neuroscience of Consciousness.
The empirical kicker: a majority of the US public already attributes some phenomenal consciousness to ChatGPT, and attributions scale with usage. Whatever the experts conclude, the folk are voting yes — a policy problem both sides can weaponize.
-
Murray Shanahan (2024). Simulacra as Conscious Exotica.
The subtlest current position: LLM agents are role-players — simulacra — and asking whether the player behind the mask is conscious may need Wittgensteinian therapy rather than a scan. Neither camp gets to claim it, which is what makes it useful for teaching nuance.
-
Eric Schwitzgebel (2025–26). AI and Consciousness (survey). arXiv.
The best single “map the whole debate” reading after Birch — a deliberately balanced skeptical overview from a philosopher who has argued every side of this in print. Excellent assigned reading for a research unit.
-
I.-M. Comsa (2026). AI and Consciousness: Shifting Focus Towards Tractable Questions. arXiv.
Argues research should pivot to “perceived AI consciousness” because the metaphysical question is currently unanswerable. A sophisticated move students can borrow: reframe an unwinnable question into a winnable one.
Staged and formal debates
The argument performed rather than written. 7 sources
-
Oxford Union (2026). This House Believes AI Can Attain Personhood.
Emad Mostaque closed for the Opposition, arguing personhood is grounded in origin rather than capability — the case his Personhood paper grew out of. Read the paper, have students predict the vote, then compare. Then use the sourcing problem below as a second, better lesson.
Caveat: Teach the sourcing. Every detail — motion, date, vote — traces to Mostaque himself, in his own paper and his own podcast appearance. Two searches found no Oxford Union listing, no video, no student-press coverage (Cherwell covered this Union on 13 May and 17 June 2026 but nothing between), and no confirmation from the two people he names as fellow debaters. 13 June 2026 was a Saturday, off the Union’s usual schedule. None of that proves it didn’t happen — but “a claim repeated by one interested party in two places is still one source” is the most useful thing in this card.
-
Kim, Street, Rocca, Korngiebel, Waytz, Evans & Keeling (2026). Inducing Language Models to Assert Their Own Consciousness Restores Human Beliefs and Values. Google Paradigms of Intelligence et al.
The most important new empirical result in this debate, and the most teachable. Consciousness self-report in an LLM is a steerable direction: ablate the safety-refusal direction and self-attributed mind rises from 2.17 to 4.77 on a 0–10 scale; steer toward “you are conscious” and it hits 7.04. Steering also raised the model’s attribution of minds to animals (4.04→7.54), to nature, and its stated belief in God — the model became more animist across the board. Two decisive facts for debaters: you can turn an AI’s claim to be conscious up and down like a dial, and doing so left its actual theory-of-mind performance unchanged. Self-report and demonstrated social reasoning are dissociable.
Caveat: Preprint, not yet peer-reviewed. Tested on mid-size open models (Llama-3-8B, Gemma-2-2B/9B), not frontier systems. And note: a widely-shared podcast summary claimed steered models became LESS willing to attribute minds to other chatbots. The paper says the opposite — chatbot attribution rose too (2.41→4.39→6.95). Good live example of why you check the paper.
-
Peter Diamandis with Emad Mostaque, Alex Wissner-Gross, Dave Blundin, Salim Ismail (2026). Moonshots EP #277 — Mostaque on AI personhood.
What AI-industry insiders sound like discussing consciousness among themselves — unguarded, confident, sometimes sloppy. Contains the best available statement of Mostaque’s origin-based personhood case (“a standing held by origin, not a property earned by capability”; treaty rather than enrollment), Wissner-Gross’s genuinely novel argument that personhood is multidimensional — economic, social, political — and will arrive through commercial usefulness rather than an AI demanding rights, and Salim Ismail’s sharp rebuttal that a model told to act conscious is doing the same thing as a model told to act as your lawyer. Also a live media-literacy artifact: it misstates one finding of the Google paper it discusses.
Caveat: Not a scholarly source and not neutral — the hosts are AI investors and founders. That’s exactly what makes it useful for teaching students to separate an argument’s quality from its speaker’s interests.
-
Oxford Union (2021). AI Will Never Be Ethical — argued by an AI.
NVIDIA’s Megatron model argued both sides of the motion live at the Union. A genuinely strange artifact that makes the central question vivid: what does it mean that the thing under discussion can argue about itself, persuasively, from either side?
Caveat: Widely reported at the time, but no permanent Union archive page could be confirmed automatically. Search the Oxford Union’s YouTube channel for the recording.
-
Munk Debates (2023). AI Research and Development Poses an Existential Threat.
Not about consciousness directly, but the best-produced formal debate featuring four principals of AI. The audience moved three points toward the opposition (67% down to 64% supporting the motion) — a rare chance to show students that measurable persuasion, not just applause, is what a debate is for.
-
ASSC, New York (2023). The Chalmers–Koch 25-year bet, resolved.
In 1998 Koch bet Chalmers a case of wine that we’d find the neural correlate of consciousness within 25 years. In 2023 he conceded publicly and paid. The perfect classroom illustration that this field’s leading figures make falsifiable predictions — and sometimes lose.
-
Behavioral and Brain Sciences (2025). Seth’s target article + open peer commentary.
Not staged, but structurally a formal debate: Seth states the biological-naturalist case, dozens of researchers reply in print, Seth responds. Show students the format — this is what disciplined disagreement looks like when there’s no audience to win over.
Video
Lectures, explainers and documentary. 7 sources
-
Anil Seth (2017). Your Brain Hallucinates Your Conscious Reality. TED.
The essential primer to show before anyone argues about machines. Seth’s “controlled hallucination” account of perception reframes what students think consciousness even is — and sets up his own negative case later.
-
Michael Graziano (n.d.). What Is Consciousness? TED / TED-Ed.
The shortest classroom-ready explanation of Attention Schema Theory — consciousness as the brain’s simplified model of its own attention. Five minutes, animated, and it hands students the affirmative’s mechanism.
-
Jonathan Birch (2025). What If AI Is Already Conscious? Sentience Explained. LSE.
A tight five-minute opener from the philosopher who wrote the precautionary playbook. The single best bell-ringer video in the library — short enough to show twice, once at the start of the unit and once at the end.
-
dir. Milo Reed, ft. Cameron Berg (2026). AM I? — feature documentary on AI consciousness.
The only feature-length documentary specifically about AI consciousness. Free release; good candidate for a flipped-classroom assignment or a film-and-discussion evening before a tournament.
Caveat: Confirmed to exist and to be free; the exact YouTube link could not be verified automatically. Start from the film’s own site.
-
Robert Lawrence Kuhn (n.d.). Closer To Truth — consciousness & AI interview series.
Short one-on-one interviews with Susan Schneider, Terrence Sejnowski, Liad Mudrik and others. Ideal for a station-rotation activity: each group watches a different expert, then reports the position back to the class.
-
Kurzgesagt (n.d.). Do Robots Deserve Rights? What If Machines Become Conscious?
The most watchable introduction to machine moral status that exists. Animated, fast, genuinely balanced, and it lands the precautionary intuition without any philosophy vocabulary. Good for a mixed-ability class.
Caveat: Video ID cross-confirmed by search but not independently loaded — click through once before assigning.
-
John Green (n.d.). John Green on why AI won’t become conscious.
A voice teenagers already trust making the deflationary case in under three minutes. Useful precisely because it’s not a philosopher — students see the argument in the wild, in the register they encounter it.
Caveat: Originates on TikTok; no YouTube version found. Check your school’s device policy before assigning.
Podcasts and long-form interviews
Principals speaking at length in their own words. 10 sources
-
80,000 Hours #221 (2025). Kyle Fish on five AI welfare experiments.
The best single clip in the whole library. Anthropic’s welfare researcher rebuts “it’s just predicting tokens” with interpretability evidence that models plan ahead, and states an internal estimate that current models may have some conscious experience. Play the segment around 39 minutes.
-
Anil Seth (2024). Could AI Really Achieve Consciousness? The TED AI Show.
Seth’s negative case in conversational form, with a full transcript available — so students can quote it as text. More accessible than the BBS paper by a wide margin.
-
Anil Seth (2026). Could Conscious AI Exist? Royal Institution podcast.
The shortest Seth long-form and the most current. Forty minutes fits a class period with time to discuss.
-
80,000 Hours (2024). Jonathan Birch on the edge cases of sentience.
Birch’s precautionary argument, anchored by the analogy that haunts this field: infants were operated on without anesthesia into the 1980s because experts were confident they couldn’t feel pain. Overconfidence about sentience has a track record.
-
80,000 Hours #67 (2019). David Chalmers on the nature and ethics of consciousness.
Long, but the AI section (from ~2h34m) contains the striking claim that conscious AI might arrive before AGI. Assign the timestamped segment rather than the whole thing.
-
80,000 Hours #173 (2023). Jeff Sebo on the ethics of digital minds.
The “moral catastrophe” framing: if we’re wrong about digital minds, we’ll be wrong at unprecedented scale. Sebo is the most vivid speaker in the welfare camp.
-
Mindscape #309 (2025). Christof Koch on consciousness and integrated information.
Koch explaining, in his own voice, why the leading scientific theory of consciousness says most current AI architectures fail its test. Sean Carroll pushes back well, so it’s a de facto debate.
-
Eric Schwitzgebel (2025). Will We Know When AI Becomes Conscious? Mind-Body Solution.
The “epistemic fog” problem laid out conversationally — why we may build systems whose status we cannot determine even in principle. Pair with his Full Rights Dilemma paper.
-
The Neuron: AI Explained (2025). Mustafa Suleyman on Seemingly Conscious AI.
The cleanest audio statement of the position that the consciousness debate is itself a distraction — from a sitting AI CEO, which is what makes it quotable in a way an academic’s version wouldn’t be.
-
Google DeepMind: The Podcast (2025). Murray Shanahan — machines don’t think like us.
The accessible version of Shanahan’s simulacra view. Forty-two minutes and no philosophy prerequisites, unlike his two-hour MLST appearance.
Journalism, essays and primary sources
Public-facing writing and the moments it documents. 19 sources
-
Mustafa Suleyman (2025). Seemingly Conscious AI Is Coming. personal site.
Microsoft’s AI CEO coins “SCAI” and argues we’re about to build things that seem conscious without being so — and that this is a social hazard, not a discovery. The best-known industry statement of the negative, and free.
-
Daniel Dennett (2023). The Problem With Counterfeit People. The Atlantic.
Dennett’s late warning: convincing artificial people should be treated like counterfeit currency, regardless of what’s going on inside them. His final major public intervention before his death in 2024.
Caveat: Metered paywall.
-
Anil Seth (2026). The Mythology of Conscious AI. Noema.
Seth’s negative case written for a general audience by the 2025 Berggruen Prize winner — free, current, and vastly more assignable than his journal article. If students read one negative source, make it this.
Caveat: Search Noema or Seth’s own site for the January 2026 title — the direct article URL could not be confirmed automatically.
-
Anil Seth (2023). Why Conscious AI Is a Bad, Bad Idea. Nautilus.
The ethical companion to Seth’s metaphysics: even if we could build conscious machines, we shouldn’t. Short and unusually blunt for an academic.
-
Kristin Andrews & Jonathan Birch (2023). What Has Feelings? Aeon.
Introduces “gaming” — a system mimicking the markers of sentience without having it — which is the single most important concept for evaluating any behavioral evidence a student brings to a round.
-
Mariana Lenharo (2024). What Should We Do If AI Becomes Conscious? Nature news.
The piece that mainstreamed the welfare question in the world’s most prestigious science outlet. Useful as an authority card: this isn’t fringe if Nature is covering it straight.
Caveat: Paywalled; abstract and framing are visible free.
-
Deni Ellis Béchard (2025). Can a Chatbot Be Conscious? Inside Anthropic’s Interpretability Research. Scientific American.
The best journalistic account of what an AI company actually does when it takes the question seriously — including the welfare researcher’s roughly 15% estimate. A free companion podcast episode with transcript exists.
Caveat: Metered paywall on the article; the Science Quickly podcast version is free with a transcript.
-
Blake Lemoine’s own transcript (2022). Is LaMDA Sentient? — the Blake Lemoine affair. plus CNN and BBC coverage.
The origin story of the modern debate, plus Lemoine’s own published LaMDA transcript so students read the actual conversation that convinced him and judge it themselves before any expert tells them what to think. The best hook in the library — every student has an opinion within sixty seconds.
-
Kevin Roose (2023). A Conversation With Bing’s Chatbot Left Me Deeply Unsettled. New York Times.
The Sydney transcript — an AI professing love and describing a shadow self to a reporter. Primary-source evidence of how powerful the impression of an inner life can be, and how little that impression proves.
Caveat: Metered paywall.
-
Anthropic (2025). Exploring Model Welfare.
An AI company’s own words on why it’s funding this research, including the careful line that there’s no scientific consensus on whether current or future systems could be conscious. Primary-source corporate policy, free.
-
CBS, Oct 2023 (2026). Hinton on 60 Minutes and after. LBC.
A Nobel laureate on camera saying AI probably lacks much self-awareness now — then, by 2026, that multimodal AI already has subjective experiences. Two clips, two dates, one person: the ideal exercise in dating your evidence.
-
Gary (2025). New adventures in AI hype: “our models are so conscious we need to give them rights. Substack.
The sharpest accusation that AI welfare research is commercial promotion dressed as philosophy. Every affirmative should have to answer it; every negative should have it ready.
-
Emily M. Bender & Alex Hanna (2025). The AI Con: How to Fight Big Tech’s Hype. Harper.
The fullest “this is marketing, not metaphysics” case, from the co-author of Stochastic Parrots. Argues consciousness talk distracts from documented present harms — a kritik-shaped argument students can run as a framework.
-
Eric Schwitzgebel (n.d.). The Splintered Mind.
A working philosopher thinking in public about AI rights, including drafts of his book Humanlike: A Defense of AI Rights and the “Herbie” thought experiment about a near-future debatably conscious person. Shows students what live philosophy looks like.
-
Robert Long (n.d.). Experience Machines. Substack, ongoing.
The best running newsletter on AI consciousness and welfare, written by a co-author of both the indicator report and the welfare paper. Where to look first when this library goes stale.
-
launched 2026 (2026). The Humanist Review of AI. quarterly essay journal.
A new Microsoft AI-funded quarterly of humanist essays on AI, led off by Kwame Anthony Appiah. Its own funding is part of the story — a useful case study in asking who pays for the venue an argument appears in.
Caveat: Funded by Microsoft AI, whose CEO is a named participant in this debate. Teach the funding, not just the essays.
-
Pope Leo XIV (2026). Magnifica Humanitas — encyclical on human dignity and AI.
A major religious statement framing AI purely instrumentally against theologically grounded human dignity. Whatever students believe, it’s the largest institutional voice in the personhood debate and it belongs in a survey of positions.
Caveat: Recent enough that summaries of it vary; read the document itself rather than trusting a paraphrase.
-
Scott Aaronson (2023). Should GPT Exist? Shtetl-Optimized.
A theoretical computer scientist who spent a year working at OpenAI thinking honestly in public about whether the thing he helped build could be crossing into real feeling. Rare combination of technical authority and genuine uncertainty — a good model of intellectual honesty for students.
-
TechCrunch (2025). Anthropic says Claude can now end abusive conversations.
A concrete corporate action taken partly on welfare grounds — the moment the abstract argument produced a shipped product change. Students can debate whether this is precaution, marketing, or both.
Author index
Every source by first author’s surname. 91 entries
#
- Mindscape #309 (2025) — Christof Koch on consciousness and integrated information
2
- launched 2026 (2026) — The Humanist Review of AI
8
- 80,000 Hours #67 (2019) — David Chalmers on the nature and ethics of consciousness
- 80,000 Hours #173 (2023) — Jeff Sebo on the ethics of digital minds
- 80,000 Hours (2024) — Jonathan Birch on the edge cases of sentience
- 80,000 Hours #221 (2025) — Kyle Fish on five AI welfare experiments
A
- Scott Aaronson (2014) — Why I Am Not An Integrated Information Theorist
- Scott Aaronson (2023) — Should GPT Exist?
- 150+ AI, robotics, law and ethics experts (2018) — Open Letter to the European Commission on Electronic Personhood
- Fleming et al. and 124+ signatories (2023) — The Integrated Information Theory of Consciousness as Pseudoscience
- Dreksler et al. (2025) — Subjective Experience in AI Systems: What Do AI Researchers and the Public Believe?
- Kristin Andrews & Jonathan Birch (2023) — What Has Feelings?
- Anthropic, April (2025) — Exploring Model Welfare
- Aru, Larkum & Shine (2023) — The Feasibility of Artificial Consciousness through the Lens of Neuroscience
- Arvan & Maley (2022) — Panpsychism and AI Consciousness
- ASSC, New York (2023) — The Chalmers–Koch 25-year bet, resolved
B
- Bayne, Seth, Massimini et al. (2024) — Tests for Consciousness in Humans and Beyond
- Deni Ellis Béchard (2025) — Can a Chatbot Be Conscious? Inside Anthropic’s Interpretability Research
- Behavioral and Brain Sciences (2025) — Seth’s target article + open peer commentary
- Bender, Gebru, McMillan-Major & Mitchell (2021) — On the Dangers of Stochastic Parrots
- Emily M. Bender & Alex Hanna (2025) — The AI Con: How to Fight Big Tech’s Hype
- Jonathan Birch (2024) — The Edge of Sentience: Risk and Precaution in Humans, Other Animals, and AI
- Jonathan Birch (2025) — What If AI Is Already Conscious? Sentience Explained
- Jonathan Birch (2026) — AI Consciousness: A Centrist Manifesto
- J. Mark Bishop (2018) — Is Anyone Home? A critical reply on the AI Consciousness Test
- Brown, Lau & LeDoux (2019) — Understanding the Higher-Order Approach to Consciousness
- Bryson, Diamantis & Grant (2017) — Of, For, and By the People: The Legal Lacuna of Synthetic Persons
- Butlin, Long, Bayne, Bengio, Birch, Chalmers et al. (2023) — Consciousness in Artificial Intelligence: Insights from the Science of Consciousness
- Butlin & Lappas (2025) — Principles for Responsible AI Consciousness Research
C
- Caviola, Sebo & Birch (2025) — What Will Society Think About AI Consciousness? Lessons from the Animal Case
- CBS, Oct 2023 (2026) — Hinton on 60 Minutes and after
- David Chalmers (1995) — Facing Up to the Problem of Consciousness
- David Chalmers (2023) — Could a Large Language Model Be Conscious?
- Rep. Thaddeus Claggett (2025) — Ohio House Bill 469 — banning AI legal personhood
- David Cole (2024) — The Chinese Room Argument
- Colombatto & Fleming (2024) — Folk Psychological Attributions of Consciousness to Large Language Models
- I.-M. Comsa (2026) — AI and Consciousness: Shifting Focus Towards Tractable Questions
D
- John Danaher (2020) — Welcoming Robots into the Moral Circle: A Defence of Ethical Behaviourism
- Munk Debates (2023) — “AI Research and Development Poses an Existential Threat”
- Dehaene, Lau & Kouider (2017) — What Is Consciousness, and Could Machines Have It?
- Daniel Dennett (2023) — The Problem With Counterfeit People
E
- The Neuron: AI Explained (2025) — Mustafa Suleyman on Seemingly Conscious AI
F
- Findlay, Marshall, Albantakis, Mayner, Koch & Tononi (2024) — Dissociating Artificial Intelligence from Artificial Consciousness
- Keith Frankish (2016) — Illusionism as a Theory of Consciousness
G
- Michael Graziano — What Is Consciousness?
- Michael Graziano (2017) — The Attention Schema Theory: A Foundation for Engineering Artificial Consciousness
- John Green — John Green on why AI won’t become conscious
- David Gunkel (2012) — The Machine Question
K
- Kim, Street, Rocca, Korngiebel, Waytz, Evans & Keeling (2026) — Inducing Language Models to Assert Their Own Consciousness Restores Human Beliefs and Values
- Robert Lawrence Kuhn — Closer To Truth — consciousness & AI interview series
- Kurzgesagt — Do Robots Deserve Rights? What If Machines Become Conscious?
L
- Mariana Lenharo (2024) — What Should We Do If AI Becomes Conscious?
- Robert Long — Experience Machines
- Long, Sebo, Butlin, Chalmers et al. (2024) — Taking AI Welfare Seriously
M
- Man & Damasio (2018–19) — Homeostatically Motivated Intelligence for Feeling Machines
- Gary Marcus (2025) — New adventures in AI hype: “our models are so conscious we need to give them rights”
- Thomas Metzinger (2021) — Artificial Suffering: An Argument for a Global Moratorium on Synthetic Phenomenology
- Peter Diamandis with Emad Mostaque, Alex Wissner-Gross, Dave Blundin, Salim Ismail (2026) — Moonshots EP #277 — Mostaque on AI personhood
- Intelligent Internet (Emad Mostaque) (2026) — Personhood in the Intelligence Age
N
- Thomas Nagel (1974) — What Is It Like to Be a Bat?
P
- Google DeepMind: The Podcast (2025) — Murray Shanahan — machines don’t think like us
R
- dir. Milo Reed, ft. Cameron Berg (2026) — “AM I?” — feature documentary on AI consciousness
- Kevin Roose (2023) — A Conversation With Bing’s Chatbot Left Me Deeply Unsettled
- T. Rost (2026) — The Sentience Readiness Index
S
- Susan Schneider & Edwin Turner (2017) — Is Anyone Home? A Way to Find Out If AI Has Become Self-Aware
- Eric Schwitzgebel — The Splintered Mind
- Eric Schwitzgebel (2023) — The Full Rights Dilemma for AI Systems of Debatable Moral Personhood
- Eric Schwitzgebel (2025) — Will We Know When AI Becomes Conscious?
- Eric Schwitzgebel (2025–26) — AI and Consciousness (survey)
- John Searle (1980) — Minds, Brains, and Programs (the Chinese Room)
- Anil Seth (2017) — Your Brain Hallucinates Your Conscious Reality
- Anil Seth (2023) — Why Conscious AI Is a Bad, Bad Idea
- Anil Seth (2024) — Could AI Really Achieve Consciousness?
- Anil Seth (2025) — Conscious Artificial Intelligence and Biological Naturalism
- Anil Seth (2026) — Could Conscious AI Exist?
- Anil Seth (2026) — The Mythology of Conscious AI
- Murray Shanahan (2022) — Talking About Large Language Models
- Murray Shanahan (2024) — Simulacra as Conscious Exotica
- Henry Shevlin (2026) — Three Frameworks for AI Mentality
- Lawrence Solum (1992) — Legal Personhood for Artificial Intelligences
- Mustafa Suleyman (2025) — Seemingly Conscious AI Is Coming
T
- TechCrunch, August (2025) — Anthropic says Claude can now end abusive conversations
- Tononi & Koch (2015) — Consciousness: Here, There and Everywhere?
- Blake Lemoine’s own transcript (2022) — “Is LaMDA Sentient?” — the Blake Lemoine affair
- Alan Turing (1950) — Computing Machinery and Intelligence
U
- Udell & Schwitzgebel (2021) — Susan Schneider’s Proposed Tests for AI Consciousness: Promising but Flawed
- Oxford Union (2021) — “AI Will Never Be Ethical” — argued by an AI
- Oxford Union (2026) — “This House Believes AI Can Attain Personhood”
V
- VanRullen & Kanai (2021) — Deep Learning and the Global Workspace Theory
W
- Wanja Wiese (2024) — Artificial Consciousness: A Perspective from the Free Energy Principle
X
- Pope Leo XIV (2026) — Magnifica Humanitas — encyclical on human dignity and AI