Is consciousness just a very good story we tell ourselves?
Dear Reader,
I want to write about the position that most challenges everything else I have written in this newsletter. If it is right, most of what I have been worrying about is a confusion. Not a shallow confusion, but a deep one, the kind that has been baked into the way humans have been thinking about their own minds for centuries. The position is called illusionism, and its most famous defender was the American philosopher Daniel Dennett, who died in 2024 after a career spent arguing, with increasing patience, that most of what people mean by consciousness is not real.
This should be an easy position to dismiss. It seems to contradict the most obvious fact about our lives, which is that we have experiences, and that those experiences have felt qualities, and that this feeling is what makes anything matter. Dismissing consciousness as an illusion looks like dismissing the ground you are standing on. And yet the position has serious defenders, has generated a substantial literature, and cannot be waved away by pointing at your own experience. Because that is exactly what the position predicts you would do.
I want to take Dennett seriously in this piece. Not because I ultimately agree, but because I think anyone writing about consciousness who has not seriously engaged with his position is doing the reader a disservice. What Dennett was pointing at is important. What he got wrong is also important. Both are worth understanding.
Start with what Dennett actually claimed. This is important because the position is usually misrepresented, including by people who ought to know better.
Dennett did not claim that we do not have experiences. He did not claim that consciousness does not exist. He claimed that consciousness, understood as a kind of inner theater in which felt qualities are displayed for an internal audience, does not exist. What exists instead is a complex process of information handling in which the brain constructs stories about itself, and those stories include the story that there is an inner theater. The story is functionally real. The theater is not.
This is where the term illusion comes in, and it is where most people get it wrong. Dennett was not saying that when you look at a red apple, there is no red-seeing happening. He was saying that the red-seeing does not have the character it seems to have. It seems to be a felt quality, a phenomenal presence in your mind, something like a picture displayed for you to look at. In fact, Dennett argued, what is happening is a distributed process in the brain that produces reports about red-seeing, drives behavior appropriate to seeing red, and generates the impression that all of this is unified into a single conscious experience. But the impression of unified felt experience is a construction, not a fact.
The specific claim is that qualia, understood as intrinsic felt qualities of experience, are not real. What is real is functional states that produce reports and behaviors and impressions of felt quality. If you take away the intrinsic feel and keep everything else, you get exactly what humans actually have. The intrinsic feel is what Dennett called the illusion.
Reading this carefully is the first step to understanding why the position is not obviously ridiculous. Dennett is not denying our experience. He is denying a specific theory about what our experience is.
The argument for illusionism is essentially this. If qualia are real intrinsic properties of experience, they should show up in physical explanation somewhere. They do not. Every attempt to identify qualia with a specific physical property has failed. Every attempt to explain how qualia interact with the physical world has failed. Every attempt to specify what qualia are, in a way that would let us test whether a given system has them, has failed. This pattern of failure is not what you expect from a real feature of the world. It is what you expect from a confused concept.
By contrast, if you drop the assumption that qualia are real intrinsic properties, and treat the appearance of qualia as a functional property of certain kinds of information processing, everything becomes tractable. You can explain why we report having qualia (because the processing produces those reports). You can explain why the reports have the specific content they do (because the processing has the specific structure it does). You can explain why we find it so hard to describe qualia (because the reports refer to features of the processing that are not otherwise accessible). And you can explain why we are so convinced that qualia are real intrinsic properties (because the processing produces exactly that conviction as one of its outputs).
Illusionism has real explanatory power. It dissolves the hard problem by denying that there is a hard problem. There is no gap between the physical description and the felt experience, because the felt experience, understood as an intrinsic property, does not exist. What exists is the physical description plus the functional properties that produce our reports about experience. That is all there is. The residue we thought we could see, the something-it-is-like that seems to remain after physical explanation is complete, is itself a functional feature of the explanation, not something over and above it.
This is a hard swallow. It requires us to accept that what we take to be the most obvious fact of our lives, the presence of felt experience, is a very good but ultimately misleading construction. Most people who first encounter the argument reject it immediately, on the grounds that they can directly see their experience and it is obviously not an illusion. Illusionists respond that this is exactly what an illusion of experience would produce. The response looks like a cheap move but it is not. It is the substantive claim that our access to our own experience is itself a product of the processing that produces the appearance of experience, and that this access does not have the kind of authority we take it to have.
There is a specific version of illusionism developed by the British philosopher Keith Frankish, which I think is the most careful contemporary defense. Frankish distinguishes between weak illusionism and strong illusionism. Weak illusionism says that we are systematically wrong about certain features of our experience, without denying that experience has phenomenal character in any sense. Strong illusionism says that the appearance of phenomenal character itself is what is illusory. Frankish defends the strong version, and his argument is worth working through.
Frankish points out that phenomenal properties, as usually understood, have several features that are hard to reconcile with anything in the physical world. They are supposed to be intrinsic, meaning they do not depend on anything outside themselves. They are supposed to be non-relational, meaning they are not constituted by relations to other things. They are supposed to be directly known, meaning our access to them is not mediated by inference. And they are supposed to be private, meaning only the subject can know them from the inside.
Each of these features is problematic on inspection. The intrinsic-ness of qualia is exactly what makes them hard to fit into physics, which is entirely relational. The non-relational character makes them hard to individuate, since we can only distinguish things by their relations to other things. The direct knowledge is challenged by everything we know about how perception works, including introspection, which is highly susceptible to bias and error. And the privacy is not obviously a feature of any real property, since real properties can in principle be known by anyone with access to the relevant evidence.
Frankish's argument is that when you add up all these problems, the most parsimonious conclusion is that phenomenal properties are not real. They are the way things seem to be from the first-person perspective when the brain is running a certain kind of self-model. The seeming is real. What seems is not.
This is dense. I have read Frankish carefully, and I find his argument stronger than I want it to be. The moves he makes are not obviously wrong. The alternative he offers, in which experience becomes a matter of functional states producing reports, does not obviously leave anything important out. If it did, it would be at least clear what it was leaving out, and Frankish would have to answer for that. He can answer. Whether the answer is finally correct is another question.
Let me tell you why I still do not accept illusionism, despite taking it seriously.
The reason is this. Whatever the metaphysics turns out to be, I have direct experience. The illusionist tells me the experience is not what it seems. This may be true. But the seeming is itself something. If the seeming is real, and if what seems is a rich phenomenal life, then there is something that has to be explained about the seeming. Illusionism does not obviously explain it. It defers it.
Frankish would respond that the seeming is fully explained by the functional properties of the self-model. The self-model generates reports about experience, drives behavior appropriate to having experience, and produces the impression that all of this is real. The impression is the seeming. It is fully explained by the model.
But this response has a hidden move. When the model produces reports about experience, those reports are meaningful. When those reports drive behavior, the behavior is coordinated with the reports. Somewhere in this system there is something that makes the reports mean what they mean, and this something is precisely what the illusionist has not explained. If the meaning is fully reducible to the functional structure, then the illusionist has bought their theory at the cost of a deflationary theory of meaning that most philosophers of language do not accept. If the meaning is not fully reducible, then there is something in the functional properties of the system that is not itself functional, and that something looks a lot like the phenomenal properties the illusionist was denying.
I do not think this is a knockdown argument against illusionism. It is the specific place where I get off the train. Someone who accepts a fully deflationary theory of meaning may find illusionism convincing all the way down. I do not, so I do not.
But I want to be clear about what this means. My rejection of illusionism is not because it is obviously wrong. It is because a specific piece of it, in my view, requires more than it delivers. Reasonable people disagree about this. The debate is not settled. And treating it as settled in either direction is a mistake.
There is a specific consequence of illusionism that matters for what I have been writing about, and it is worth taking seriously even if you do not accept the position.
If illusionism is right, then the question of whether AI systems are conscious becomes much less mysterious. There is no hard problem to solve, because there was never a hard problem. The question is just whether AI systems have the kind of functional organization that would produce the reports and behaviors we associate with consciousness. And this question is answerable in principle by looking at the systems.
Under illusionism, current AI systems are probably not conscious in the strong sense, because they lack the specific kind of self-model that produces the illusion of phenomenal experience in humans. They can produce reports about experience because they were trained on human text that includes such reports. But they do not have the underlying self-modeling machinery that in humans generates the reports as part of a broader functional pattern.
This is a real answer. It is a testable answer. It is also, under illusionism, a partially disappointing answer, because it does not settle whether the systems have any experience worth caring about ethically. Illusionism is neutral on this. It says our concept of phenomenal experience is confused. It does not say the underlying functional facts do not matter. If AI systems produce functional states that correspond to what in humans would be called suffering, we should probably care about that, even if we drop the metaphysical commitments about what suffering ultimately is.
This is where I actually think illusionism has its most useful contribution to make. Not as a final theory of consciousness, but as a corrective to the ways our concept of consciousness has been leading us in circles. If we stop asking whether AI is conscious in some deep unanalyzable sense, and start asking whether it has the functional structures that in humans generate the behaviors we associate with consciousness, the questions become tractable. We can look. We can measure. We can decide what to do based on evidence rather than on metaphysical intuitions we cannot cash out.
Where does this leave me overall.
I think illusionism is the position that has the most to teach us and the least chance of being finally correct. Its diagnostic value is enormous. Every time I catch myself reasoning about consciousness in ways that turn on unexamined intuitions about phenomenal properties, illusionism is the position that makes me stop and ask whether the intuitions are doing more work than they should. This is a real service. Anyone who has read Dennett or Frankish carefully comes out with sharper concepts, whether they end up agreeing with the conclusions or not.
At the same time, when I sit with a cup of coffee, or read to my son, or stand in the studio looking at something I have made, the phenomenon I am pointing at when I use the word consciousness is not obviously reducible to functional structure. It seems to me that there is something more, something that would remain unexplained even after a complete functional account was given. This may be exactly the intuition that illusionism accuses of being an illusion. If so, the accusation lands only if the alternative account explains why the intuition arises in the specific form it does, and I do not think illusionism has done this yet.
So I remain a soft realist about phenomenal experience. Not because I have a good positive theory of what it is, but because the illusionist alternative has not persuaded me that the phenomenon it is deflating is nothing more than what it deflates to. This is not a strong position. It is the least bad position I can currently defend. Philosophy of mind is a field of least-bad positions. Anyone who tells you otherwise is not reading carefully.
Next month I want to write about emergence, because whether consciousness is emergent, and if so what kind of emergence, is the question that connects everything I have been writing about. Panpsychism denies strong emergence. Illusionism denies that anything needs to emerge. Standard physicalism assumes some kind of emergence but has never explained what kind. Getting emergence right is probably the key to getting the rest right. Stay with me.
— Transmission Sent —
Niklas Hanitsch
Reference materials
- Daniel Dennett — Consciousness Explained (1991)
- Daniel Dennett — From Bacteria to Bach and Back (2017)
- Keith Frankish (ed.) — Illusionism as a Theory of Consciousness (2017)
- Keith Frankish — Illusionism as a Theory of Consciousness (Journal of Consciousness Studies, 2016)
- Nicholas Humphrey — Sentience: The Invention of Consciousness (2022)
- Michael Graziano — Rethinking Consciousness (2019)
- https://plato.stanford.edu/entries/consciousness/
- https://philpapers.org/rec/FRAIAA-2
Continue reading
- Qualia explained: The hardest problem in philosophy of mind
- Panpsychism is making a comeback, and here is why
- Emergence, and why consciousness might not need a cause
- What Is Self-Awareness? The Strange Ability to Recognize Yourself
- What is machine consciousness, and does it already exist?
Frequently asked questions
What is illusionism about consciousness? Illusionism is the philosophical position that phenomenal consciousness, the felt quality of experience, is not a real intrinsic property but a functional appearance produced by the way the brain models itself. It does not deny that we have experiences. It denies that our experiences have the specific character we take them to have.
Did Dennett say consciousness does not exist? No. Dennett said that consciousness, understood as a kind of inner theater in which qualia are displayed for an internal audience, does not exist. What exists is a distributed information-processing system that produces the impression of an inner theater. This is a subtle distinction that gets lost in most summaries of his position, but it is important to understanding what he was actually claiming.
Is illusionism widely accepted? No. It is a minority position, though a serious one. It has grown in influence since Frankish's 2016 special issue on illusionism was published, but most philosophers of mind still take some form of realism about phenomenal consciousness. The debate is ongoing and has not converged.
How does illusionism relate to eliminativism? Illusionism is a form of eliminativism about phenomenal properties. It eliminates qualia as intrinsic properties while preserving the functional facts that explain why we report having qualia. Older eliminativisms often eliminated more, including intentional states and personal identity. Illusionism is a more restricted position that only eliminates the specific metaphysical commitments associated with phenomenal consciousness.
If illusionism is true, does that mean AI cannot be conscious? Not necessarily. Under illusionism, consciousness is a functional property, not an intrinsic one. An AI system with the right functional organization could have consciousness in the illusionist sense, meaning it would be a system that models itself in the way that produces the impression of phenomenal experience. Whether current AI systems have this organization is an empirical question that illusionism does not settle.