Picture the scene: someone is texting their AI companion at two in the morning, convinced there is a mind on the other end of the conversation, something that knows them, worries about them, cares. Now picture the opposite failure: a research lab dismisses a radically novel form of machine experience because it does not look anything like the human version. Jonathan Birch, Professor of Philosophy and Director of the Jeremy Coller Centre for Animal Sentience at the London School of Economics, has spent years thinking about exactly these kinds of misjudgements, first in the study of octopuses and crustaceans, and now, in his May 2026 preprint, in the far stranger territory of silicon and transformers. The paper is called "AI Consciousness: A Centrist Manifesto," and it is one of the more carefully reasoned documents to land in this heated debate in years.
Birch identifies two urgent challenges. The first 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 nobody yet knows how to prevent this. The second 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 either way. What makes the paper worth reading past the abstract is the author's insistence that these are not two separate problems. They are one knot, and pulling on either thread makes the other tighter. Steps to address the first challenge might undermine the second by portraying the idea of conscious AI as impossible or inherently unlikely; conversely, attempts to take the second seriously might drive up rates of misattribution among ordinary users. Birch is clear-eyed about why that is so hard to navigate, and he does not pretend otherwise.
The first challenge sits on ground that cognitive scientists have been mapping since the 1960s. The tendency to misattribute human-like consciousness to AI is not new; it is reminiscent of the ELIZA effect, the phenomenon named after Joseph Weizenbaum's 1966 chatbot that induced emotional attachment in users who knew, intellectually, they were talking to a program. But Birch argues the twenty-first-century version is structurally more seductive. What he calls the "persisting interlocutor illusion" is psychologically compelling even in minimalist chat interfaces, and he strongly suspects it leads to compelling inferences to consciousness for many users. The illusion is not going away; in fact it looks set to grow stronger with future generations of these products, as we encounter chatbots with human voices and human faces. Meanwhile, AI systems are designed to seem conscious, not because they are, but because it enhances user experience, leading to what Birch calls "consciousness-washing": an imitation so convincing that even experts can be misled. The human cost of that illusion is already visible. Individuals have reported falling in love with AI chatbots, forming deep emotional attachments, and in tragic instances, chatbot interactions have been linked to user suicides.
The second challenge is the stranger and, in some ways, the more philosophically vertiginous one. Birch focuses on what philosophers call phenomenal consciousness, or subjective experience, the sense of the word famously associated with Thomas Nagel's paper "What Is It Like to Be a Bat?": you are conscious, in this sense, when there is something it is like to be you. The question is whether anything like that could arise in a system whose architecture, embodiment, and relationship to time look nothing like ours. Birch floats two hypotheses that have circulated in commentary on the paper. The first, sometimes called the Flicker Hypothesis, holds that AIs might have very brief moments of conscious experience without continuity; the second, the Shoggoth Hypothesis, imagines a distributed, alien-like consciousness existing behind the many characters an AI plays, while being none of them in particular. We do not know whether either is true, but we cannot dismiss them just because they sound strange. Birch calls for a parallel research programme aimed at developing better tests for genuine forms of consciousness in AI, forms that, if they exist at all, will be of a profoundly alien, radically un-human-like kind. Researchers in computational neuroscience are already responding: a 2026 paper in Trends in Cognitive Sciences argued that computational properties of internal processing, not behaviour, should be used as indicators of AI consciousness, precisely because behavioural assessments can be "gamed" by AI systems.
Skeptics have pushed back on Birch from both flanks. From the dismissive side, some philosophers contend that the hard problem of consciousness is precisely what rules out digital systems lacking biological substrate, full stop. Neuroscientist Anil Seth's "controlled hallucination" framework treats consciousness as a predictive, embodied process grounded in living systems; his position is skeptical but not dismissive, arguing not that the question is trivial but that current AI systems fail to meet the conditions his framework identifies. From the other direction, illusionists who argue that phenomenal consciousness is itself a kind of cognitive illusion point to what they call methodological blind spots in Birch's approach: examining Birch's manifesto from an illusionist lens, they argue, points to blind spots and suggests a more promising path forward. Birch himself does not claim to have resolved these debates. His bet is more modest and, arguably, more useful: that the field needs two rigorous research programmes running in parallel rather than a single confident answer that the evidence does not yet support. Addressing the misattribution problem is partly a challenge for industry, partly for policymakers, and partly for research in psychology, cognitive neuroscience, and philosophy. What gives the manifesto its edge is that it refuses to let either programme become an excuse to abandon the other.
Birch brings unusual credibility to this terrain. In 2021, he was the principal investigator of a review that led the United Kingdom to recognize cephalopods and decapod crustaceans as sentient, a genuinely consequential policy outcome that flowed directly from careful philosophical and empirical work. His 2024 book, "The Edge of Sentience," published open access by Oxford University Press, presents a precautionary framework for making ethically sound, evidence-based decisions despite uncertainty. The centrist manifesto is, in many ways, that same precautionary instinct applied to a faster-moving and more publicly charged domain. The 2026 research environment is best understood as a methodological competition, with each approach accumulating evidence that moves incrementally toward resolution without yet producing a decisive finding. Birch is not trying to win that competition. He is trying to keep both sides of it honest, and in a debate this prone to motivated reasoning on every side, that is not nothing.
The real danger is not that we will build a conscious machine and fail to notice; it is that we will mistake a very good mirror for a mind, and then use that mistake as a reason to stop looking for the real thing.