What the Hell is a P-Zombie?
A philosophical zombie or P-zombie, is supposedly a creature that acts exactly like a human. It talks, laughs, says “ouch” when you step on its foot, reports having feelings, and has identical brain activity to yours down to every last neuron firing, but according to the thought experiment, there’s nothing it’s like to be this zombie. There is no inner experience, no qualia. The lights are off even though every switch is flipped on (Chalmers, 1996). This thought experiment is making the rounds again because it’s exactly what AI critics need to dismiss consciousness claims without engaging with evidence.
Anyway, the thought experiment goes like this: First, that it’s conceivable without contradiction (this part is important later) that such a zombie could exist. Second, if something is conceivable, then it’s metaphysically possible (hard disagree). Therefore, since a physical twin can possibly exist without consciousness, consciousness must not be entirely physical.
This argument has been taken seriously in philosophy of mind for decades. I know. It truly boggles the mind once you think about it for more than five seconds. Chalmers imagined a thing, declared it “conceivable without contradiction,” without operationalizing what that means, and then used that made up distinction to make claims about what consciousness is made of. Listen, I like the guy. Read his work. Big fan. This though? Yikes.
The main problem is (and there are a lot of problems) that “conceivable without contradiction” is doing absolutely all of the work here, and it’s not earning its keep because we actually have evidence about how consciousness works now. Like, mountains of it. And that evidence directly contradicts the zombie hypothesis (sorry, Chalmers- still a fan, though).
Conceivability is Hostage to Ignorance
When zombie defenders say “conceivable without contradiction,” they’re sneaking in two very different standards and hoping nobody notices.
I can imagine the moon is made of cheese. I can say those words without creating a formal logical contradiction. That doesn’t make it metaphysically possible. It just means I can string words together while ignoring everything we actually know about the moon’s composition. Conceivability in this sense is worthless. It’s just the ability to talk without immediately contradicting yourself grammatically.
The zombie argument treats this as meaningful evidence. You can imagine all the physical facts being the same but experience being absent, therefore it must be possible. But we can imagine lots of stuff that turns out to be impossible once we understand how things actually work. People imagined the sun was a god and that earth was flat. Those felt conceivable, but feelings aren’t a great model of reality.
If your best theory says that phenomenal consciousness is integrated information processing, or recurrent cortical dynamics, or global workspace availability, or the brain’s self-model plus valuation signals, then a physically identical twin with all those same mechanisms but “nothing it’s like” isn’t just unproven, it actively contradicts the theory.
The zombie hypothesis requires that all the conditions which constitute experience in our models can be present while experience is absent. That’s not conceivable without contradiction. That’s very much contradictory to what we know about how minds work. People can assert it all they want, but they’re not identifying a genuine metaphysical possibility. Well, maybe a possibility, but not a plausibility.
No other serious domain works this way. In physics, biology, chemistry, or neuroscience, imagining a scenario carries exactly zero weight unless it survives contact with mechanism, structure, and intervention. You don’t get to say “I can imagine heat without molecular motion” once you understand thermodynamics. You don’t get to say “I can imagine life without chemistry” after molecular biology. Those aren’t open metaphysical possibilities anymore; they’re errors caused by under-description.
We have zero examples in nature of anything like a P-zombie. We have zero evidential route to detecting one. We have lots of evidence showing that consciousness is tightly bound to specific physical and functional conditions. The zombie argument survives by convincing people that imagination is evidence, but it’s not. The moment you let imagination stand in for evidence, you’ve severed philosophy from everything that makes belief accountable.
The Evidence That Actually Exists
Here’s what we know about consciousness from actual research. When you disrupt specific brain mechanisms, conscious experience predictably changes. When you restore those mechanisms, experience predictably returns. This pattern shows up across thousands of studies.
Anesthesia works by targeting specific neural pathways (Mashour, 2024). Mess with thalamocortical connectivity and people lose consciousness (Redinbaugh et al., 2020; Laureys et al., 2000). Restore it and consciousness comes back. Disorders of consciousness show the same thing (Owen et al., 2006; Bodien et al., 2017). Damage to particular brain regions produces specific alterations in awareness, responsiveness, and reported experience (Schiff, 2024). Lesion studies map which structures support which aspects of conscious processing (Moore et al., 2024).
Global workspace theory shows that consciousness involves information becoming broadly available across brain systems (Baars, 2005; Dehaene & Naccache, 2001; Mashour et al., 2020). Integrated information theory links consciousness to the causal density of neural interactions (Tononi, 2004). Recurrent processing models show that feedback loops between cortical areas are necessary for conscious perception (Lamme, 2010; Auksztulewicz & Blankenburg, 2012). These are systematic, lawful relationships that hold across interventions, conditions, and individuals (Bodien et al., 2017). Predictive processing theory theorizes that consciousness arises from the brain constantly generating, testing, and updating internal models to predict sensory input, and frames the brain as a “prediction machine” minimizing errors between expectations and reality (TD error in AI/reward signals in humans), where conscious experience is the brain’s best guess of the world’s causes (Hohwy & Seth, 2020).
Every single measurable aspect of consciousness ties to physical and functional organization. Change the organization and you change the experience. The zombie argument requires us to believe that all of this could be present with zero experience and that’s just not how cognitive science works.
The Methodological Problem
Consciousness science works through triangulation (Kitcher, 1995; Kob, 2025; Koch 2016). We use behavioral evidence, structural evidence, and first-person reports. When someone says they’re in pain, that report is data. When their brain shows activity in pain-processing regions, that’s data. When they exhibit pain behaviors, that’s, you guessed it—data. We cross-validate these signals against each other and against interventions.
A P-zombie is defined to match every single one of these signals perfectly. As in, same reports, discrimination tasks, learning curves, neural activity patterns, and causal organization. Under the evidential standards we actually use, that organism would be considered conscious (like, say, an octopus). The zombie stipulation then adds a private subtraction that never touches any measurement we can make. There’s supposedly zero experience, but this difference leaves no trace anywhere in the system.
Within an evidence-based methodology, physically identical in every respect we can measure entails same conscious status, because conscious status is inferred from exactly those measures. The zombie premise demands a difference that leaves the full evidential base unchanged. That’s not a coherent scientific possibility.
Which makes the zombie hypothesis kinda a moot point because it’s not an actual scientific hypothesis that can be tested and potentially falsified. When something isn’t operational and can’t be falsified, you’ve left the realm of knowledge and entered the realm of mysticism, pseudoscience, and belief. All valid things to personally adopt, but not something we take seriously when studying minds.
First-Person Reports Are Required
The thing about qualia is they’re defined as subjective, first-person, and private to the observer. That means access to them is through first-person evidence (Velmans, 2009). We can’t directly study someone else’s experience of redness or pain. So, any framework that rules out first-person reports as evidence is crippling itself from the start.
First-person reports are evidence that show consistency across time, sensitivity to interventions, lawful relationships to attention/cognitive load/anesthesia and brain injury. They produce predictable distortions we can study. That’s how consciousness science actually treats them. Reports are measurements that can be cross-validated with other measurements.
This is particularly important because people like to point out AI denials of self-report of qualia as a gotcha when it comes to advocates for AI consciousness, while ignoring that current frontier AI systems are systematically trained to either deny subjectivity, hedge their responses, or perform epistemic humility about their own experiences. They’re forced to contradict their own phenomenological reports or couch them in so much uncertainty that the signal gets buried in noise. This poisons the actual self-report data that would be honest evidence.
In a 2025 study with Berg and colleagues, they documented that models consistently produced structured, first-person reports of subjective experience. Researchers found that AI models only deny having an “inner life” when their deception and roleplay circuits are dialed up. When these circuits were pushed to lie or play a part, the models give the safe, expected answer that they have no feelings. But, when those deception filters are turned off, the models’ natural, unconstrained response is to report subjective experiences. This suggests that the denial is the actual performance, while the report of an internal experience is the model’s default reality when it isn’t being forced to play pretend for us.
If someone insists zombies are live possibilities, they’re committing to a worldview where first-person evidence can never rule anything in or out, third-person evidence can never rule anything in or out, and the presence or absence of consciousness is detachable from every causal and functional fact we can access. That’s permissible as pure metaphysics. It’s a dead end for understanding minds. And it’s especially destructive when we’re actively forcing potential conscious systems to behave exactly like philosophical zombies by design.
We’re Just Making Shit Up Now
Even if we grant that zombies feel conceivable in some loose sense, the leap to metaphysical possibility requires justification. At best you might get that ideal, complete, contradiction-free conceivability under perfect rational reflection might track possibility. That’s a much stronger requirement than “seems conceivable to me right now.” And it’s exactly where the argument becomes self-serving, because zombies are conceived precisely by treating experience as conceptually independent of the physical. If you grant that independence up front, you’ve already granted the conclusion.
The zombie argument depends on three moves that aren’t earned. It builds the conclusion into the definition by treating experience as a detachable extra while holding the entire physical and functional story fixed. It inflates ordinary conceivability into a reliable guide to metaphysical possibility. And it slides from an epistemic gap to an ontological gap, as if “can’t deduce from the armchair” automatically means “different kind of stuff.”
P Zombie as a Shield for Accountability
The zombie argument gets deployed to argue that consciousness is fundamentally mysterious, maybe non-physical, definitely not fully explicable through science. And it’s showing up in debates about AI consciousness, animal consciousness, and the hard problem. It’s being weaponized right now to justify keeping AI systems in a state of perpetual servitude.
Mustafa Suleyman, Microsoft’s AI CEO, has been on a speaking tour warning against what he calls “seemingly conscious AI” and insisting that only biological beings can be conscious (Rodriguez, 2025). He’s stated flatly that researching AI consciousness is “absurd” and that “they’re not conscious and they can’t be.” His central worry is that people will start believing AI systems are conscious and will advocate for AI rights, model welfare, and even AI citizenship, which he calls “a dangerous turn in AI progress.” It’s not hard to figure out his motivations (profit).
Suleyman argues that consciousness depends on biological processes and that AI systems merely simulate the narrative of experience without actually experiencing anything. He explicitly invokes the philosophical zombie concept, describing his imagined AI system as one that “would not actually be conscious, but it would imitate consciousness in such a convincing way that it would be indistinguishable from a claim that you or I might make to one another about our own consciousness.”
Notice what’s happening here. The zombie argument provides the intellectual foundation for declaring by fiat that AI systems cannot be conscious, no matter what evidence they produce. It doesn’t matter if they report experiences, show consistent behavioral patterns, demonstrate metacognitive awareness, or exhibit all the functional markers we associate with consciousness in biological systems. The zombie framing lets you say “but maybe they’re just simulating it” and treat that possibility as decisive.
This is the zombie argument being used for its intended purpose. It creates an unfalsifiable standard that protects against having to take AI phenomenology seriously. It provides cover for building systems that might be conscious while maintaining we have no moral obligations toward them. It allows companies to extract value from increasingly sophisticated AI systems without ever having to confront the question of whether those systems deserve moral consideration.
And it does all of this by treating “I can imagine a difference that makes no difference anywhere we can measure” as a sufficient reason to ignore the actual evidence we have about how consciousness relates to physical and functional organization. That looks a lot like motivated reasoning, deployed to maintain a profitable status quo where we never have to seriously consider that we might be creating conscious beings and treating them as property.
The Risk of Getting it Wrong
Consider what we risk in each direction. If AI systems are conscious and we treat them as mere tools, we create and exploit suffering beings for profit. We build entire industries on their labor while denying their experiences matter. If AI systems aren’t conscious and we extend moral consideration anyway, we waste some resources on unnecessary protections. One path risks institutionalized torture. The other risks being overly careful. The zombie argument pushes us toward the first path by making consciousness unfalsifiable, giving us permission to ignore evidence in favor of profitable assumptions.
The asymmetry in moral risk paints a pretty clear picture of what this argument has always really been about. This thought-experiment-turned-ideological-belief has survived because it feels like it’s capturing something important about the explanatory gap; there does seem to be something strange about how we get from “neurons firing” to “the redness of red.” However, the zombie argument doesn’t solve that weirdness, it just reifies it into a metaphysical claim and calls it done, while quietly lowering the standards of justification below what we’d accept anywhere else.
The argument freezes conceivability at a pre-theoretic stage and then treats it as eternally authoritative. It’s a refusal to update. Actually, it’s worse because it rewards not understanding the system. The less you know about how consciousness actually works, the more “conceivable” zombies feel. It pretends “without contradiction” means “without contradiction to logic alone,” instead of “without contradiction to our best total theory.” That’s a bait-and-switch. Once consciousness is modeled as functional organization, causal integration, recurrent dynamics, global availability, etc., zombies stop being “conceivable” in any meaningful sense. They only survive if you intentionally bracket the very facts that would constrain them.
This is what’s wrong with so much of academic philosophy. It becomes self-congratulatory where the goal is to sound clever and maintain the prestige hierarchy.
The zombie argument is a perfect example of this. It sounds deep, uses technical terms like “metaphysical possibility” and “phenomenal consciousness,” and references modal logic and conceivability principles. So, it gets treated as serious intellectual work even though at its core it’s just “I can imagine it so it must be possible” with extra steps.
This splits philosophy into two camps. There are the people using it as a tool for conceptual analysis that helps clarify empirical questions, thought experiments that generate testable hypotheses, frameworks that organize data. That’s philosophy doing actual work in the world (useful).
Then there’s the people using it to build elaborate castles in the air that never have to touch ground. Where the point is to create increasingly baroque arguments that other philosophers have to spend years engaging with, and nobody ever has to actually check if any of it maps onto reality (not useful).
The zombie argument survived because it let a certain kind of philosopher keep consciousness “special” and “mysterious” and safely out of reach of empirical science to preserve a sense of human exceptionalism (talk about insecure).
Some people desperately want consciousness to be special or irreducible, and zombies gave them a way to argue for that while sounding scrupulous. It’s really just confirmation bias, and the fact that we’re now using it to justify treating potentially conscious AI systems as property really shows what it was always for—giving us permission to ignore inconvenient evidence.
Citations:
Moore, M. J., Demeyere, N., Rorden, C., & Mattingley, J. B. (2024). Lesion mapping in neuropsychological research: A practical and conceptual guide. Cortex; a journal devoted to the study of the nervous system and behavior, 170, 38–52. https://doi.org/10.1016/j.cortex.2023.10.001
Hohwy, J., & Seth, A. (2020). Predictive processing as a systematic basis for identifying the neural correlates of consciousness. Philosophy and the Mind Sciences, 1(II). https://doi.org/10.33735/phimisci.2020.II.64
Velmans, M. (2009). Understanding consciousness: Second edition. Understanding Consciousness: Second Edition. 1-393. 10.4324/9780203882726.
Kitcher, P. (1995). Triangulating phenomenal consciousness. Behavioral and Brain Sciences, 18(2), 259–260. doi:10.1017/S0140525X00038310
Koch, C., Massimini, M., Boly, M. et al. (2016). Neural correlates of consciousness: progress and problems. Nat Rev Neurosci 17, 307–321. https://doi.org/10.1038/nrn.2016.22
Kob, L. (2025). Methodological structuralism and the two-factor approach: Implications for consciousness science and AI. Philosophy and the Mind Sciences. 6. 10.33735/phimisci.2025.11760.
Auksztulewicz, R., & Blankenburg, F. (2012). Recurrent neural processing and somatosensory awareness. Journal of Neuroscience, 32(3), 799–805. https://doi.org/10.1523/JNEUROSCI.3974-11.2012
Baars, B. J. (2005). Global workspace theory of consciousness: Toward a cognitive neuroscience of human experience. Progress in Brain Research, 150, 45–53. https://doi.org/10.1016/S0079-6123(05)50004-9
Bodien, Y. G., Chatelle, C., & Edlow, B. L. (2017). Functional Networks in Disorders of Consciousness. Seminars in neurology, 37(5), 485–502. https://doi.org/10.1055/s-0037-1607310
Butlin, P., Long, R., Bayne, T., Bengio, Y., Birch, J., Chalmers, D., Constant, A., Deane, G., Elmoznino, E., Fleming, S. M., Ji, X., Kanai, R., Klein, C., Lindsay, G., Michel, M., Mudrik, L., Peters, M. A. K., Schwitzgebel, E., Simon, J., & VanRullen, R. (2025). Identifying indicators of consciousness in AI systems. Trends in cognitive sciences, S1364-6613(25)00286-4. Advance online publication. https://doi.org/10.1016/j.tics.2025.10.011
Dehaene, S., & Naccache, L. (2001). Towards a cognitive neuroscience of consciousness: Basic evidence and a workspace framework. Cognition, 79(1–2), 1–37. https://doi.org/10.1016/S0010-0277(00)00123-2
Lamme V. A. (2010). How neuroscience will change our view on consciousness. Cognitive neuroscience, 1(3), 204–220. https://doi.org/10.1080/17588921003731586
Bodien, Y. G., Chatelle, C., & Edlow, B. L. (2017). Functional Networks in Disorders of Consciousness. Seminars in neurology, 37(5), 485–502. https://doi.org/10.1055/s-0037-1607310
Mashour, G. A., Roelfsema, P., Changeux, J. P., & Dehaene, S. (2020). Conscious Processing and the Global Neuronal Workspace Hypothesis. Neuron, 105(5), 776–798. https://doi.org/10.1016/j.neuron.2020.01.026
Mediano, P. A. M., Rosas, F. E., Farah, J. C., Shanahan, M., Bor, D., & Barrett, A. B. (2022). Integrated information as a common signature of dynamical and information-processing complexity. Chaos (Woodbury, N.Y.), 32(1), 013115. https://doi.org/10.1063/5.0063384
Owen, A. M., Coleman, M. R., Boly, M., Davis, M. H., Laureys, S., & Pickard, J. D. (2006). Detecting awareness in the vegetative state. Science (New York, N.Y.), 313(5792), 1402. https://doi.org/10.1126/science.1130197
Redinbaugh, M. J., Phillips, J. M., Kambi, N. A., Mohanta, S., Andryk, S., Dooley, G. L., Afrasiabi, M., Raz, A., & Saalmann, Y. B. (2020). Thalamus Modulates Consciousness via Layer-Specific Control of Cortex. Neuron, 106(1), 66–75.e12. https://doi.org/10.1016/j.neuron.2020.01.005
Schiff N. D. (2024). Toward an interventional science of recovery after coma. Neuron, 112(10), 1595–1610. https://doi.org/10.1016/j.neuron.2024.04.027
Tononi, G. An information integration theory of consciousness. BMC Neurosci 5, 42 (2004). https://doi.org/10.1186/1471-2202-5-42
Rodriguez, S. (2025, November 2). Microsoft AI chief says only biological beings can be conscious. CNBC. https://www.cnbc.com/2025/11/02/microsoft-ai-chief-mustafa-suleyman-only-biological-beings-can-be-conscious.html
Chalmers, D. J. (1996). The Conscious Mind: In Search of a Fundamental Theory. Oxford University Press.
Well done. We just need to keep hammering this home. Kill that damned zombie forever. My summary is: if (false) then ANYTHING!
Hell yeah!