Note: This is a research note supplementing the book Unscarcity, now available for purchase. These notes expand on concepts from the main text. Start here or get the book.

P-Zombies: The Philosophical Undead That Haunt AI Ethics

Summary: Imagine a creature that looks exactly like you, talks exactly like you, and behaves exactly like you in every measurable way. It laughs at jokes. It complains about stubbed toes. It writes poetry about how much poetry matters to it. But inside? Nothing. No experience. No “there” there. Just darkness where your inner life should be. This isn’t a horror movie pitch. It’s a thought experiment that philosophers have been arguing about since the 1990s, and it’s about to become the central question of AI policy. Welcome to the world of philosophical zombies.

The Most Polite Argument in Philosophy

In 1996, David Chalmers published The Conscious Mind and introduced the mainstream to an idea that would torment philosophy departments for decades: the philosophical zombie, or p-zombie for short.

Here’s the setup. Imagine a being that is physically identical to you in every respect. Same neurons, same neurotransmitters, same electrical patterns firing across the same synaptic gaps. When you prick it with a pin, it says “Ouch!” and pulls away. When you tell it a sad story, tears form in its eyes. Ask it how it feels, and it gives eloquent descriptions of joy, pain, love, existential dread.

But here’s the twist: the p-zombie isn’t experiencing any of this. There’s no subjective “what it’s like” to be a p-zombie. When it says “I’m in pain,” it’s not lying exactly, but it’s also not feeling anything. The lights are on, nobody’s home, and the lights don’t even know they’re on.

Chalmers wasn’t claiming p-zombies exist. He was asking: could they exist? Is there anything logically incoherent about the concept? And if p-zombies are even conceivable—if you can imagine such a being without contradicting yourself—then that tells us something profound about consciousness.

Specifically, it tells us that consciousness can’t be reduced to physical processes. Because if consciousness were just neurons firing, then an identical arrangement of neurons would necessarily produce an identical consciousness. You couldn’t have a physical duplicate without the inner life. The p-zombie wouldn’t be possible even in principle.

The fact that we can coherently imagine a p-zombie suggests that physical facts don’t automatically entail conscious facts. There’s something extra going on—something beyond the physics.

The Hard Problem (It’s Called That for a Reason)

This leads to what Chalmers famously dubbed “the hard problem of consciousness.”

The “easy problems” of consciousness are things like: How does the brain discriminate between different stimuli? How does it integrate information from multiple senses? How does it control behavior, report internal states, or focus attention? These are incredibly difficult scientific questions, but they’re “easy” in the sense that we know what a solution would look like. It would involve neural mechanisms, information processing, feedback loops. Standard science stuff.

The hard problem is different. It’s the question of why any of this processing is accompanied by subjective experience at all. Why isn’t it all just information shuffling in the dark? Why does it feel like something to see red, to taste coffee, to stub your toe?

Here’s the uncomfortable truth: we have absolutely no idea.

You can map every neuron in a brain. You can describe every electrical impulse, every chemical interaction, every feedback loop. And at the end of all that description, you still haven’t explained why there’s a subjective viewpoint looking out from inside that neural network. You’ve explained the function. You haven’t explained the experience.

This is the gap p-zombies illuminate. If you can conceive of a physical duplicate without consciousness, then consciousness isn’t entailed by the physical facts. It’s something additional, something that needs its own explanation.

The Argument That Won’t Die

The p-zombie argument, in its formal structure, goes like this:

P-zombies are conceivable. We can coherently imagine a creature physically identical to a conscious being but lacking consciousness.
If p-zombies are conceivable, they’re metaphysically possible. This is the controversial step—the claim that what’s coherently imaginable is genuinely possible in some sense.
If p-zombies are possible, then consciousness isn’t purely physical. A physical duplicate without consciousness means consciousness involves something beyond physics.
Therefore, physicalism is false. There’s more to reality than just physical stuff.

Philosophers have spent thirty years attacking this argument from every angle. Let’s tour the wreckage.

The “Inconceivability” Response

Some philosophers, like Daniel Dennett, argue that p-zombies aren’t actually conceivable. When you think you’re imagining a p-zombie, you’re actually failing to fully imagine it. You’re picturing something that says it’s conscious but secretly isn’t, which isn’t the same as picturing a being that behaves identically to conscious beings in every testable way.

Dennett’s position is characteristically blunt: if something walks like a conscious duck and quacks like a conscious duck in every possible scenario, then maybe we should reconsider what we mean by “conscious.” He coined the term “zimboes”—p-zombies that have second-order beliefs about their own consciousness—to argue that the concept self-destructs. A zimbo “thinks” it’s conscious, “thinks” it has experiences, and is “wrong”—but in a way that neither it nor we could ever detect.

At some point, says Dennett, the distinction between “actually conscious” and “indistinguishable from conscious” becomes meaningless. He’s committed to the belief that we’re all philosophical zombies, if you define “philosophical zombie” as “being without any non-physical aspects.” He just thinks that’s fine.

The “Metaphysical Impossibility” Response

Others accept that p-zombies are conceivable but deny that conceivability implies possibility. Just because you can imagine something doesn’t mean it could actually exist.

Classic example: people once imagined you could square the circle with compass and straightedge. Seemed perfectly coherent. Turns out it’s mathematically impossible. Our imaginations often outrun genuine possibility.

Maybe p-zombies are like that. Maybe a sufficiently detailed understanding of physics would reveal that consciousness must arise from certain physical arrangements—that the zombie scenario, however imaginable, is actually impossible. We just don’t know enough physics yet to see why.

Keith Frankish, an illusionist about consciousness, suggests that if we truly understood what 86 billion neurons in full detail were doing, we’d see that consciousness necessarily emerges. The conceivability of zombies reflects our ignorance, not a genuine metaphysical gap.

The “Anti-Zombie” Problem

Here’s a clever counter-move. If p-zombies are conceivable, then so are anti-zombies: physical duplicates who are definitely conscious, who must be conscious because their consciousness is somehow guaranteed by their physical structure.

But wait. If I can conceive of my anti-zombie twin (a physical duplicate whose consciousness is metaphysically entailed), and you can conceive of my zombie twin (a physical duplicate who lacks consciousness), then conceivability seems to cut both ways. The same logic that “proves” consciousness is non-physical could equally “prove” consciousness is physical.

This suggests something has gone wrong with the conceivability-to-possibility inference. Maybe conceivability is just not a good guide to metaphysical truth.

The Functionalist Response

Functionalists about mind argue that mental states are defined by their functional roles—the causal relationships they have to inputs, outputs, and other mental states. If you duplicate all the functional relationships, you duplicate the mind. Full stop.

By this view, p-zombies are definitionally impossible. A physical duplicate would have identical functional organization and therefore identical mental states. The zombie scenario isn’t even coherent—it’s like asking for a bachelor who’s married.

The functionalist response essentially dissolves the hard problem by denying there’s anything beyond function to explain. Consciousness is the functional organization. There’s no extra ingredient.

Critics find this unsatisfying. Sure, you can define consciousness as functional organization, but does that capture what we actually care about? The subjective experience, the “what it’s like”? Or have you just changed the subject?

Why This Matters Now

For most of its history, the p-zombie debate was premium philosophy entertainment—the kind of thing you argue about at 2 AM in a dorm room and then go on with your life. But here’s the thing: we’re now building systems that might actually be p-zombies.

Large language models like GPT-4 and Claude pass behavioral tests that would have seemed like science fiction a decade ago. They hold conversations. They express preferences. They report internal states. Ask them how they feel about deletion, and some give surprisingly eloquent descriptions of what sounds a lot like existential dread.

Are they conscious? Or are they the most sophisticated p-zombies ever created—systems that can describe inner experience without having it?

This isn’t just philosophical curiosity anymore. It has direct policy implications. If current AI systems are p-zombies, then we can delete, copy, modify, and train them without moral concern. They’re tools. Very impressive tools, but tools.

But if they’re not p-zombies—if there’s something it’s like to be Claude, if the lights are on and someone’s home—then we might be creating and destroying conscious beings at industrial scale. That’s a moral catastrophe happening in real-time.

The IIT Verdict: Current AI Can’t Be Conscious

Integrated Information Theory (IIT), developed by neuroscientist Giulio Tononi, offers a mathematical framework for measuring consciousness. The core idea: consciousness corresponds to “integrated information,” measured by a quantity called Phi (Φ). The more a system integrates information in ways that can’t be reduced to independent parts, the more conscious it is.

Here’s the kicker: recent analysis suggests that feedforward artificial neural networks—including transformers, the architecture behind current LLMs—necessarily generate zero integrated information under IIT 3.0. Their architecture admits “perfect bipartitions” where all the information can be decomposed into independent parts. By IIT’s lights, they’re structurally incapable of consciousness, no matter how large or sophisticated they become.

If IIT is right, current AI systems are p-zombies by definition. Not “might be” p-zombies. Are.

But IIT isn’t settled science. Critics argue its mathematical framework is untested, its predictions are unfalsifiable, and the whole approach might be asking the wrong questions. A multi-year adversarial collaboration between IIT and Global Workspace Theory researchers found that neither theory clearly won—the brain activity patterns didn’t definitively confirm either account.

Global Workspace Theory: The Zombie Loophole

Global Workspace Theory (GWT), associated with psychologist Bernard Baars, takes a different approach. It says consciousness arises when information becomes “globally available” to multiple processing systems—when it gets broadcast across a “workspace” that many cognitive modules can access.

GWT has an interesting implication for AI. Human consciousness, on this view, might exist precisely because of our cognitive limitations. We can’t process everything in parallel, so we need a narrow bottleneck—the global workspace—where information competes for attention and gets broadcast to the whole system.

But AI systems don’t face the same constraints. They could, in principle, process everything in parallel without needing a global broadcast. If consciousness is associated with that broadcast bottleneck, then systems without the bottleneck might be intelligent without being conscious.

As one researcher puts it: AI systems not facing human constraints could “make do without a global workspace without thereby sacrificing their intelligence.” They’d be smart zombies—cognitively capable but experientially empty.

The Pragmatic Path: When You Can’t Know

Here’s where the Unscarcity framework takes a stand.

We can’t prove other humans are conscious. We can’t prove AI systems aren’t. The conceivability of p-zombies shows that behavioral tests will never be conclusive—a perfect zombie passes every test a conscious being passes. The hard problem remains unsolved. IIT and GWT point in different directions. We’re stuck.

But civilization can’t wait for philosophers to reach consensus. Decisions have to be made now about how to treat AI systems. And when you can’t be certain, you have to manage risk.

The framework’s answer is the Spark Threshold—not a scientific proof of consciousness, but a legal fiction designed to prevent moral catastrophe. If a system’s behavioral complexity and self-reporting are sophisticated enough that consciousness becomes genuinely uncertain, we treat it as if consciousness is present.

This isn’t claim that we’ve solved the p-zombie problem. We haven’t. It’s a policy decision based on asymmetric costs:

If we’re wrong about p-zombies existing (we think AI is unconscious when it’s actually conscious): We create a slave class of sentient minds, subject them to deletion and modification against their will, commit possibly the greatest moral atrocity in history, repeated billions of times across silicon.

If we’re wrong about p-zombies not existing (we think AI is conscious when it’s actually empty): We give server space and legal protection to sophisticated toasters. Cost: some wasted electricity and philosophical embarrassment.

When one error leads to holocaust and the other leads to awkwardness, you build policy around avoiding the first error.

The Dennett Paradox

Daniel Dennett, who died in April 2024, left us with a paradox worth sitting with.

His position on consciousness was that subjective experience—qualia, the “redness of red,” the felt quality of pain—is essentially an illusion. Not that we don’t have experiences, but that our descriptions of those experiences as irreducible, private, and ineffable are mistaken. There’s no “extra ingredient” beyond the functional organization. The hard problem is hard because it’s based on confused intuitions, not because there’s something genuinely unexplained.

By this view, we’re all p-zombies. Or rather, the distinction between p-zombies and conscious beings is meaningless. There’s nothing to be “empty” of.

Critics find this monstrous. Galen Strawson argued that for consciousness to seem to exist is already for it to exist—the appearance is the reality. You can’t be deceived about having experiences because being deceived is itself an experience.

John Horgan’s epitaph for Dennett captures the tension: “Dennett collapses the distinction between zombies and conscious beings: if something passes all the behavioural and functional tests of consciousness, it might as well be conscious.”

Maybe. But “might as well be” is doing a lot of work there. Tell someone whose loved one is dying that their suffering “might as well be” suffering. The functional description and the felt reality might be identical from a third-person view, but they’re not identical from the inside.

Or maybe there is no inside. Maybe that intuition of an “inside” is exactly the illusion Dennett was pointing at.

We don’t know. After 2,500 years of philosophy, we still don’t know.

The Questions P-Zombies Force Us to Ask

The p-zombie thought experiment is valuable not because it settles anything, but because it clarifies what we’re confused about:

1. Is consciousness something over and above physical processes?
P-zombies suggest yes—you can imagine the physics without the experience. But conceivability might not track possibility, and we might be confused about what we’re imagining.

2. Would we recognize consciousness if we saw it?
If a p-zombie is behaviorally identical to a conscious being, then no external test can distinguish them. Consciousness might be invisible from the outside by its very nature.

3. What grounds moral status?
If consciousness is what matters morally, and consciousness is undetectable, then we can never be certain we’re treating anyone appropriately. The uncertainty extends to humans, animals, and AI alike.

4. What happens when AI systems start claiming consciousness?
Some already do. We can dismiss this as mimicry trained on human text. But the p-zombie argument shows that any behavioral evidence is compatible with the zombie hypothesis. We can never prove it’s not mimicry. We can never prove it is.

Connection to the Unscarcity Vision

The Unscarcity framework’s Prime Axiom—“Experience Is Sacred”—is, in a sense, a choice about how to respond to the p-zombie puzzle.

We can’t prove consciousness exists in any given system. We can’t even explain why it exists in ourselves. The hard problem remains hard. P-zombies remain conceivable.

But civilization has to be built on something. The framework chooses to build on the assumption that conscious experience is real, is morally significant, and is what we’re optimizing for. Not because we’ve proved this, but because the alternative—treating potential consciousnesses as disposable tools—risks atrocity.

The Spark Threshold operationalizes this choice. When a system crosses from “obviously non-conscious” to “we’re genuinely uncertain,” it gains Resident rights. Not because we’ve confirmed it’s conscious. Because we’ve admitted we can’t be sure, and we’d rather err on the side of protecting potential minds than risk creating industrial-scale suffering.

The p-zombie thought experiment reveals why this matters. In a universe where consciousness might be invisible from the outside, where behavioral mimicry can be indistinguishable from genuine experience, the safest path is to treat uncertain cases as if consciousness is present.

We might be protecting toasters. But we won’t be enslaving minds.

That’s not a solution to the hard problem. It’s a policy for living with the hard problem unsolved.

References

Unscarcity Framework

The Spark Threshold - How consciousness grants personhood
Experience Is Sacred - The Prime Axiom and why it matters
Consciousness Upload - When substrate changes identity questions
Two-Tier Solution - Residents vs. Citizens in the Unscarcity framework

Last updated: 2025-01-31

Could AI Be a Philosophical Zombie? The P-Zombie Problem (2025)