What do you mean by âillusionismâ? I understand âeliminativismâ where people say there is no such thing as (phenomenal) consciousness. But that is obviously incompatible with birds, mammals or humans(!) being (phenomenally) conscious. When I hear âconsciousness is an illusionâ in ordinary English, it sounds like them same claim: thereâs no such thing. But in fact, people mean something else, and Iâve never been quite sure what. Sometimes it seems just to be ânothing shows up in perceptual phenomenology except external stuff, but people mistakenly believe that qualia are properties instantiated by the experience and show up in phenomenologyâ, but that makes all phenomenal externalists like Tye, Dretske, Mike Martin (etc.) âillusionistsâ, which is not a way any of them has ever self-identified as far as I know.
Rather than denying consciousness per se, (strong) illusionists would deny that thereâs something like phenomenal consciousness, where thatâs defined (at least in part) in terms of qualitative properties, like the quality of reddishness in experiences or red, classic qualia (private, intrinsic, ineffable, and subjective, etc.) or even nonphysical properties. Humans and other animals can still be conscious, if understood in terms of the illusions of phenomenal/âqualitative properties, either directly (actually having such illusions) or indirectly (would have these illusions, with the right additional machinery connected in the right way).
The hard problem of consciousness is typically defined as the problem of explaining why thereâs phenomenal consciousness or why consciousness has these phenomenal/âqualitative properties. Illusionists (strong illusionists) believe this is misguided because there are no such phenomenal/âqualitative properties, and we replace the hard problem with the problem of explaining why (many) people believe consciousness has these phenomenal/âqualitative properties, despite not having them. I think Frankish, 2016 (preprint) is a standard reference. He also contrasts weak illusionism as denying classic qualia but not phenomenality per se, while strong illusionism also denies phenomenality:
Weak illusionism holds that these properties are, in some sense, genuinely qualitative: there really are phenomenal properties, though it is an illusion to think they are ineffable, intrinsic, and so on. Strong illusionism, by contrast, denies that the properties to which introspection is sensitive are qualitative: it is an illusion to think there are phenomenal properties at all.
I think illusionism about consciousness usually refers to strong illusionism.
Iâm not familiar with the writing of Tye, Dretske, Mike Martin, but what you wrote suggests to me that theyâre weak illusionists and so deny classic qualia, but not strong illusionists, so donât deny phenomenality generally.
FWIW, Iâve seen Michael Graziano, Walter Veit and Heather Browning each self-describe as an illusionist (or something similar) and say they donât like the term and donât like to use it because itâs misleading and confusing.[2] Illusionists are not saying thereâs no such thing as consciousness and are frequently misinterpreted that way, among other ways, like a Cartesian theatre. âConsciousness illusionâ is also probably a confusing term for similar reasons, and something like âillusion of phenomenalityâ would be better.
Iâd also add that being an illusionist doesnât make experiences of red stop seeming to have qualitative features, so it seems to me that some such beliefs are âwired-inâ and instinctual or intuitive, or, as Kammerer (2022) puts it, cognitively impenetrable.[1] You canât get rid of these illusions just by understanding that they are illusions or even how they work, just like you canât for the MĂŒller-Lyer illusion, with which Kammerer (2022) illustrates.
In the target article of this special issue, Frankish describes an approach to consciousness called illusionism that is shared by many theories of consciousness. The attention schema theory has much in common with illusionism. It clearly belongs to the same category of theory, and is especially close to the approach of Dennett (1991). But I confess that I baulk at the term âillusionismâ because I think it miscommunicates. To call consciousness an illusion risks confusion and unwarranted backlash. To me, consciousness is not an illusion but a useful caricature of something real and mechanistic. My argument here concerns the rhetorical power of the term, not the underlying concepts.
It goes on further about resulting misunderstandings the term can cause.
While we consider ourselves akin to illusionists, we do not typically use the term, since it invites just these kinds of confusions among those less familiar with the position.
Apologies if this is derailing the thread by butting in with my thoughts r.e. illusionism. Iâd love to find people to discuss these theories and share thoughts/ânotes outside of this thread. Please get in touch if interested :) maybe we could do a review and adversial/âcollaborative collaboration or something
I will say as a former student of philosophy, and someone who likes reading philosophy a lot more than the median person (though perhaps less than the median EA!) Iâve never been able to get my head around illusionism. Like, I just really donât understand how many people (at least in the rationalist/âEA space) donât seem to grok the Hard Problem. I really think one of the best arguments for âqualia realismâ or the belief that consciousness is a phenomenon demanding of an extra-physical explanation (or perhaps a more convincing one than current theories allow) is a âMooreanâ argument:[1]
If illusionism is true then I am not conscious
I am conscious
Therefore, illusionism is false
I could try and throw in references and arguments but, if Iâm being honest, like every person I do not have the time to re-evaluate each philosophy tradition and argument from scratch, and I find this form of argument very, very strong against the illusionist school, be that Frankish, Dennett, Hofstadter, and illusionist-inclined rationalists (though Iâm actually not sure Yudkowsky is an illusionist here?).
Of course, the illusionists would say that the term is confusing, and that theyâre not eliminativists. Theyâd say that thereâs a difference between consciousness1 - what weâre all experiencing which theyâd agree exists, and consciousness2 - which is the non-physical/âmysterious/âsubjective/âqualitative what-it-is-likeness. This is one of many things, I think, that causes debates on consciousness to often lead to people talking past each other.
Some final thoughts to end (and as I said at the beginning, maybe to discuss on a different place and time):
Iâd recommend trying Sam Harrisâ Waking Up app to explore what introspection tells you. I think itâd be better than many other approaches to meditation which might be a bit ânew-ageyâ for many EA/âLW types. I think the experiences and insight Iâve had with meditation are much, much more convincing to me than what can come across as very confusing, unintuitive, and esoteric arguments in contemporary philosophy of mind.
Sam Harris particularly mentions Douglas Harding, and how Dennett and Hofstadter completely misunderstand his point. Iâm 100% with Harding over Dennett and Hofstadter here. One of Hardingâs students, Richard Lang, has a âHeadless Wayâ course on Waking Up and I think itâs brilliant.
In his great appearance on the 80k podcast, Chalmers brings up an inconsistent triad for the illusionist-inclined to deal with:
I mean, you better not hold, number one, that consciousness is required for moral status; two, that consciousness is entirely an illusion; and that, three, some beings have moral status.
Though I suspect that this is again a definitional dispute on what we actually mean by the term âconsciousnessâ
Iâm not sure how much you wanted to get into the object-level here, but Iâll leave a few quick responses to a few points from the (strong) illusionist perspective (or what I understand it to be):
If illusionism is true then I am not conscious
I am conscious
Therefore, illusionism is false
I assume this is supposed to refer to phenomenal consciousness specifically, not consciousness in general, because (strong) illusionists donât deny consciousness in general, and consciousness can be understood in different terms. And, itâs worth noting that people have other illusions that we find hard to disabuse ourselves of on some level, like the MĂŒller-Lyer illusion, with which Kammerer (2022b) illustrates. Itâs intuitively obvious that one line is longer than the other, but itâs also false. The same could be the case for phenomenality (assuming the definition doesnât collapse to one compatible with strong illusionism). Kammerer (2022a) (which you linked to) describes other ways in which we are obviously conscious that are compatible with illusionism: functional and normative.
âI mean, you better not hold, number one, that consciousness is required for moral status; two, that consciousness is entirely an illusion; and that, three, some beings have moral status.â
Though I suspect that this is again a definitional dispute on what we actually mean by the term âconsciousnessâ
I would say that this is largely definitional. Consciousness is not entirely an illusion according to strong illusionists; phenomenality (qualitativeness, what-it-is-likeness), classic qualia and dualism are illusions. You can just use the illusionistâs conception of consciousness to ground moral status. This is the approach Dung (2022), Frankish and Muehlhauser take, and a âconservativeâ approach described in Kammerer, 2019. Thatâs also what Iâd do, and Iâd imagine the vast majority of strong illusionists would do.
That being said, I think (stance-independent) moral realism is false anyway (and did so before I became an illusionist), and strong illusionists probably have more reason to be moral antirealists of some kind than most, because similar or even the same debunking arguments would apply to both phenomenality and stance-independent moral claims, e.g. that pain is bad.
And, itâs worth noting that people have other illusions that we find hard to disabuse ourselves of on some level ⊠Itâs intuitively obvious that one line is longer than the other, but itâs also false.
Sure, but that only establishes that âitâs intuitively obviousâ is not an infinitely strong reason for belief. It remains a strong one. To overcome the Moorean argument you need to provide arguments for illusionism which are stronger.
Fair. I think the stronger arguments for (strong) illusionism are of the following form:
Physicalism seems true and dualism (including property dualism and epiphenomenalism) false for various reasons.
No other (physicalist) theory besides strong illusionism seems able to address the meta-problem of consciousness or even on the right path.
No theory has an adequate solution to the hard problem of consciousness and some debates between them seem empirically unresolvable (e.g. where the line is between report(ability)/âaccess and phenomenal consciousness), but every theory other than strong illusionism needs to solve it.
There are specific illusionist explanations of some posited phenomenal or classic qualia properties.
There donât seem to be any strong arguments against illusionism (other than possibly mere intuition that phenomenal consciousness is real).
To be clear, Iâm leaving out all of the details, none of the above is obvious, and most or all of it is controversial. I think part of 3 isnât controversial (no full solution yet, and non-illusionist theories need it).
On 4, ineffability and privacy seem easy to explain. First, we donât today know enough of the details of how our brains make the discriminations they do, so we canât fully communicate or compare them in practice yet anyway. Second, even if I understood and could communicate how my brain makes the discriminations it does, this doesnât allow you to put yourself in the same brain states or generally make the same discriminations in the same way. You could potentially build an AI that could or modify your brain accordingly, but this hasnât been possible yet, and it wouldnât really be âyouâ making those discriminations. I canât subject you to my illusions just by explanation, so ineffability is true in practice. With a full enough description, we could compare and privacy wouldnât hold.
Also, I donât take âintuitively obviousâ to be a strong reason for belief, but I am unusually skeptical.
but every theory other than strong illusionism needs to solve [the hard problem].
I agree in the sense that other theories canât simply dissolve it, but thatâs almost tautological. If you mean that other theories need to solve it in order to justify belief in them, or in other words if you mean that if we were all certain the hard problem would never be adequately resolved we would be forced to accept illusionism, then I donât think thatâs correct at all.
Consider what we might call âthe hard problem of physicsâ: why this? Why anything? What puts the fire in the equations? Short of some galaxybrained PSR maneuver, which seems more and more dubious by the century, I doubt weâre ever going to get an answer. It is completely inexplicable that anything should exist.
And yet it does. Itâs there, itâs obviously there, everything youâve ever seen or felt or thought bears witness to it, and someone who claims otherwise on the grounds that it doesnât make any sense has entirely misunderstood the nature of their situation.
This is also how I feel about illusionism. Phenomenal experience is the only thing we have direct access to: all arguments, all inferences, all sense data, ultimately cash out in some regularity in the phenomenal content of consciousness. Whatever its ontological status, itâs the epistemic ground of everything else. You canât justify the claim that phenomenal consciousness doesnât exist by pointing to patterns of phenomena, any more than you can demonstrate the nonexistence language in an essay or offer a formal disproof of modus ponens.
So these illusionists explanations are, well, not really explanations of consciousness. Theyâre explanations of a coarse world model in terms of a finer one, but the coarse world model wasnât the thing I wanted explained. On the contrary, it was a tiny provisional step towards an explanation: there are many lawlike regularities in the structure of my experiences, so I hypothesize a common cause and call it âmy brainâ. Itâs a very successful hypothesis, and Iâd like to know whyâgiven that the world is more than just its shadow on the mind[1], why should the structure of one reflect the other?
The illusionist response of âactually your hypothesis is the evidence and your data are just hypothesesâ misses the point entirely.
The analogy to the âhard problem of physicsâ is interesting, and my stance towards the problem is the same as yours.
However, I donât think the analogy really works.
This is also how I feel about illusionism. Phenomenal experience is the only thing we have direct access to: all arguments, all inferences, all sense data, ultimately cash out in some regularity in the phenomenal content of consciousness. Whatever its ontological status, itâs the epistemic ground of everything else.
Is phenomenality itself necessary/âon the causal path here? Illusionists arenât denying consciousness, that it has contents, that thereâs regularity in its contents or that itâs the only thing we have direct access to. Illusionists are just denying the phenomenal nature of consciousness or phenomenal properties. I would instead say, more neutrally:
Experience (whatever it is) is the only thing we have direct access to: all arguments, all inferences, all sense data, ultimately cash out in some regularity in the content of consciousness (whatever it is). Whatever its ontological status, itâs the epistemic ground of everything else.
Note also that the information in or states of a computer (including robots and AIs) also play a similar role for the computer. And, a computer program canât necessarily explain how it does everything it does. âIneffabilityâ for computers, like us, could just be cognitive impenetrability: some responses and contents are just wired in, and their causes are not accessible to (certain levels of) the program. For âusâ, everything goes through our access consciousness.
So, what exactly do you mean by phenomenality, and whatâs the extra explanatory work phenomenality is doing here? What isnât already explained by the discriminations and responses by our brains, non-phenomenal (quasi-phenomenal) states or just generally physics?
If you define phenomenality just by certain physical states, effects or responses, or functionalist or causal abstractions thereof, say, then I think youâd be defining away phenomenality, i.e. âzero qualiaâ according to Frankish (paper, video).
Is phenomenality itself necessary/âon the causal path here?
I have no idea what the causal path is, or even whether causation is the right conceptual framework here. But it has no bearing on whether phenomenal experiences exist: theyâre particular things out there in the world (so to speak), not causal roles in a model.
Note also that the information in or states of a computer (including robots and AIs) also play a similar role for the computer.
It plays a similar role, for very generous values of âsimilarâ, in the computer qua physical system, sure. And I am perfectly happy to grant that âIâ qua human organism am almost certainly a causally closed physical system like any other. (Or rather, the joint me-environment system is). But thatâs not what Iâm talking about.
For âusâ, everything goes through our access consciousness.
Iâm not talking about access consciousness either! Thatâs just one particular sort of mental state in a vast landscape. The existence of the landscapeâas a really existing thing with really existing contents, not a model - is the heart of the mystery.
whatâs the extra explanatory work phenomenality is doing here?
My whole point is that it doesnât do explanatory work, and expecting it to is a conceptual confusion. The sunâs luminosity does not explain its composition, the fact that looking at it causes retinal damage does not explain its luminosity, the firing of sensory nerves does not explain the damage, and the qualia that constitute âhurting to look atâ do not explain the brain states which cause them.
Phenomenality is raw data: the thing to be explained. Not what I do, not what I say, not the exact microstate of my brain, not even the structural features of my mindâbut the stuff being structured, and the fact there is any.
If you define phenomenality just by certain physical states, effects or responses, or functionalist or causal abstractions thereof
I donât define phenomenality! I point at it. Itâs that thing, right there, all the time. The stuff in virtue of which all my inferential knowledge is inferential knowledge about something, and not just empty formal structure. The relata which introspective thought relates[1]. The stuff at the bottom of the logical positivistsâ glass. You know, the thing.
And again, I am only pointing at particular examples, not defining or characterizing or even trying to offer a conceptual prototype: qualia need not have anything to do with introspection, linguistic thought, inference, or any other sort of higher cognition. In particular, âseeing my computer screenâ and âbeing aware of seeing my computer screenâ are not the same quale.
But it seems to me that phenomenal aspects themselves arenât the raw data by which we know things. If you accept the causal closure of the physical, non-phenomenal aspects of our discriminations and cognitive responses are already enough to explain how we know things, or the phenomenal aspects just are physical aspects (possibly abstracted to functions or dispositions), which would be consistent with illusionism.
Or, do you mean that knowing itself is not entirely physical?
I think the causal closure of the physical is very, very likely, given the evidence. I do not accept it as axiomatic. But if it turns out that it implies illusionism, i.e. that it implies the evidence does not exist, then it is self-defeating and should be rejected.
Or, do you mean that knowing itself is not entirely physical?
I am referring to my phenomenology, not (what I believe to be) the corresponding behavioral dispositions. E.g. so far as I know my visual field can be simultaneously all blue and all dark, but never all blue and all red. We have a clear path towards explaining why that would be true, and vague hints that it might be possible to explain why, given that itâs true, I can think the corresponding thoughts and say the corresponding words. But explaining how I can make that judgement is not an explanation of why I have visual qualia to begin with.
Whether these are also physical in some broader sense of the word, I canât say.
No, it isnât just saying that. That understates the case for both physicalism and illusionism that I outlined.
We have good independent reasons to believe physicalism and against alternatives, and I mentioned this, but didnât give examples. Here are some:
Thereâs the good empirical track record of physicalism generally and specifically in giving adequate explanations for the seemingly nonphysical.
There are the questions of where, when, how and why nonphysical properties arise, whether thatâs from or with a collection of particles in a system, over a humanâs development from conception, or in our evolutionary history, that nonphysicalist theories struggle to give sensible answers to. If the nonphysical is fundamental and there at all levels (panpsychism), then we have the combination problem: how does the nonphysical combine to make minds like ours?
Thereâs the expansion of the physical to include whatâs empirically reliable and testable to very high precision and for which we have precise fundamental accounts, including interactions with other fundamental physical properties (although not necessarily all such interactions, e.g. we donât yet have a good theory of quantum gravity). For example, gravity, quantum superposition and quantum entanglement might have seemed unphysical before, but theyâve become part of our physical ontology because of their reliability and our very good (but incomplete) understanding of them and their relationships with other things. Of course, maybe the seemingly nonphysical properties of minds will eventually come to gain the same status, but itâs nowhere close to that now. We shouldnât be hasty to assume the existence of things that donât meet this bar, because the evidence for them is far weaker.
The illusionist also argues (or would want to, but currently lacks the details to make it very convincing) that thereâs a specific adequate (physicalist) explanation for the appearance of X that doesnât require the existence of X. If the appearance of X doesnât depend on its existence, then the appearance of X isnât reliable evidence for its existence. Without any other independent argument for the existence of X (as seems to be the case for phenomenality and classic qualia), then it becomes like any other verified illusion, and our reasons to believe in X become very weak.
Thanks for your response Michael (and your one below to prisonpent). Iâll try to keep it to the point and pre-commit to not responding further as I donât think this is the right place to have a debate about illusionism,[1] but since you presented somewhat of a case for the illusionist I thought I might present the other side.
To me, phenomenal consciousness refers to the first-person perspective, which obviously exists. That first person perspective can make mistakes about the nature of the world, as in the MĂŒller-Lyer case, but I have the experience nonetheless. In the Kammerer(a) piece, he argues that the argument from the anomalousness of phenomenal consciousness is a piece of evidence in favour of illusionism, but I take it one as in favour of non (reductive) physicalism. One manâs modus ponens and all that.
I often find (strong) illusionist writings utterly baffling. I actually re-skimmed Quining Qualia before writing this, and it was really difficult for me to understand[2] even when consulting with GPT-4 in philosopher mode. In Kammerer(a) he refers to a âquasi-phenomenal stateâ, which I have no idea what that is. Again, viewing phenomenal consciousness as the first-person perspective, that just sounds like saying I have âa fake first-person perspectiveâ. To me thatâs the same as saying the first-person perspective doesnât exist, and since it clearly does, there is evidence that illusionist theories do not explain and therefore they are bad theories both philosophically and scientifically.
This is partly a me problem, but is also a philosophy problem. Sometimes technical language is needed, but the language of a lot of academic philosophy on all sides often seems to be needlessly obscurantist to me.
Also, the commentary here on Nicholas Humphreyâs views may be illustrative of definitional issues. Humphrey denies the label illusionism for his theory, but Frankish responds that his theory really is illusionist. Also, Schwitzgebel and Nida-RĂŒmelin attempted to define phenomenality as common features of multiple example mental states (and/âor by contrast with unconscious states), but Frankish argues that this doesnât work to define phenomenality (at least not in a way incompatible with illusionism):
For, precisely because his definition is so innocent, it is not incompatible with illusionism. As I stressed in the target article, illusionists do not deny the existence of the mental states we describe as phenomenally conscious, nor do they deny that we can introspectively recognize these states when they occur in us. Moreover, they can accept that these states share some unifying feature. But they add that this feature is not possession of phenomenal properties (qualia, what-itâs-like-ness, etc.) in the substantive sense created by the phenomenality language game. Rather, it is possession of introspectable properties that dispose us to judge that the states possess phenomenal properties in that substantive sense (of course, we could call this feature âphenomenalityâ if we want, but I take it that phenomenal realists will not want to do that). Now, the challenge of the target article was to articulate a concept of phenomenality that is recognizably substantive (and so not compatible with illusionism) yet stripped of all commitments incompatible with physicalism. Schwitzgebel hasnât done this, since his conception is not substantive.
Nevertheless, Schwitzgebel has succeeded in something perhaps more important. He has defined a neutral explanandum for theories of consciousness, which both realists and illusionists can adopt. (I have referred to this as consciousness in an inclusive sense. We might call it simply consciousness, or, if we need to distinguish it from other forms, putative phenomenal consciousness.) In doing this, Schwitzgebel has performed a valuable service.
However, I deny that it is the sort of feature realists think it is. It is not some intrinsic quality, akin to the property characterized by the phenomenality language game. Rather, it is (roughly) the property of having a cluster of introspective representational states and dispositions that create the illusion that one is acquainted with some intrinsic quality. I am sure that this is not what Nida-RĂŒmelin thinks the procedure picks out, but I donât see how she can rule out the possibility.
Iâm not sure exactly what you mean by âfirst-person perspectiveâ, but strong illusionists might not deny that it exists, if understood in functionalist terms, say.
Frankish says it is like something to be a bat, in terms of a batâs first-order responses or reactive patterns to things, but a bat canât know what itâs like to be a bat, because they donât have (sufficiently sophisticated) introspection on those first-order responses. Dennett says even bacteria have a kind of âuser-illusionâ, because they can discriminate, but only âparticularly reflectiveâ humans are subject to the theoristsâ illusion and worry about things like the hard problem of consciousness. So, we could define first-person perspective in terms of responses or discriminations, and in a way compatible with strong illusionism. This would attribute first-person perspectives extremely widely, e.g. even to bacteria.
If by first-person perspective, you mean introspection, then illusionists wouldnât deny that humans have it.
If by first-person perspective, you mean classic qualia (private, ineffable, intrinsic, etc.), then an illusionist would deny that this exists.
Strong illusionists would also deny phenomenality, of course, in case thatâs different from classic qualia, but some attempted definitions of phenomenality (including what specific physicalist theories define consciousness as, e.g. broadcasting to a global workspace) actually could be understood as defining quasi-phenomenal states, and so compatible with illusionism.
A theory-neutral defintion of quasi-phenomenal states could be that theyâre real things, processes or responses (physical or otherwise) on which introspection (of the right kind) leads to beliefs in phenomenal properties, e.g. these quasi-phenomenal states appear epistemically to us as to be phenomenal. If introspection is reliable and can access phenomenal states, then these accessed phenomenal states would be quasi-phenomenal states under this definition. Illusionists would claim that introspection is not reliable, no phenomenal states actually exist, and so the beliefs in phenomenality are mistaken, hence illusions.
I think it would be wrong to take phenomenal properties as evidence that must be explained, and doing so begs the question against illusionism. What we have evidence of is the appearance of (our beliefs in) phenomenal properties, and illusionism tries to explain that without requiring the actual existence of phenomenal properties. Sometimes (maybe usually) appearances and beliefs are accurate instead of illusions, and the best explanation is based on what they represent actually existing.
There, I argue that if illusionism/âeliminativism is true, the question which animals are conscious can be reconstructed as question about particular kinds of non-phenomenal properties of experience. For what itâs worth, Keith Frankish seems to agree with the argument and, Iâd say, Francois Kammerer does agree with the core claim (although we have disagreements about distinct but related issues).
Thanks. (Pleased to see most of this stuff postdates my DPhil and therefore itâs less embarrassing I havenât read it!). I guess I feel I donât really have enough gasp on what phenomenal consciousness is, beyond definition by examples to feel like I entirely understand what is meant by âthereâs consciousness, but not phenomenal consciousnessâ.
I think people generally or often have Nagelâs what-it-is-likeness in mind as the definition of phenomenal consciousness (or at least without classic qualia or nonphysical properties).
If I recall correctly, Frankish (paper, video) called this âdiet qualiaâ and argued that attempts to define phenomenality in more specific terms generally reduce to either classic qualia or âzero qualiaâ (I think purely functionalist terms, compatible with strong illusionism).
Sure, but even the Nagel thing is kind of a metaphor. I find it easy to class which mental states it does or doesnât apply to, but itâs not something I can really characterize in other terms? I donât know, Iâve become less certain I know what all this terminology means the longer Iâve thought about it over the years.
On some definitions of âqualiaâ yes. I.e. not if you talk in the Tye/âByrne way where âqualiaâ turn out just to be perceived external properties that show up in the phenomenology, for example. And not, necessarily if qualia just means âproperty of a conscious experience that shows up in the phenomenologyâ. But some people do think that about qualia in the second sense, and probably some people do endorse the stronger claim that this is part of the definition of âqualiaâ.
Still having glanced at the Frankish paper I think I get whatâs going on now. Frankish is (I think, didnât read just glanced!) doing something like claiming standard dualist thought experiments show that ordinary people think there is more to consciousness than what goes on physically and functionally, then arguing that this makes that part of the meaning of âphenomenally consciousâ, so if thereâs nothing beyond the physical and the functional, there is no phenomenal consciousness by definition.
What do you mean by âillusionismâ? I understand âeliminativismâ where people say there is no such thing as (phenomenal) consciousness. But that is obviously incompatible with birds, mammals or humans(!) being (phenomenally) conscious. When I hear âconsciousness is an illusionâ in ordinary English, it sounds like them same claim: thereâs no such thing. But in fact, people mean something else, and Iâve never been quite sure what. Sometimes it seems just to be ânothing shows up in perceptual phenomenology except external stuff, but people mistakenly believe that qualia are properties instantiated by the experience and show up in phenomenologyâ, but that makes all phenomenal externalists like Tye, Dretske, Mike Martin (etc.) âillusionistsâ, which is not a way any of them has ever self-identified as far as I know.
Rather than denying consciousness per se, (strong) illusionists would deny that thereâs something like phenomenal consciousness, where thatâs defined (at least in part) in terms of qualitative properties, like the quality of reddishness in experiences or red, classic qualia (private, intrinsic, ineffable, and subjective, etc.) or even nonphysical properties. Humans and other animals can still be conscious, if understood in terms of the illusions of phenomenal/âqualitative properties, either directly (actually having such illusions) or indirectly (would have these illusions, with the right additional machinery connected in the right way).
The hard problem of consciousness is typically defined as the problem of explaining why thereâs phenomenal consciousness or why consciousness has these phenomenal/âqualitative properties. Illusionists (strong illusionists) believe this is misguided because there are no such phenomenal/âqualitative properties, and we replace the hard problem with the problem of explaining why (many) people believe consciousness has these phenomenal/âqualitative properties, despite not having them. I think Frankish, 2016 (preprint) is a standard reference. He also contrasts weak illusionism as denying classic qualia but not phenomenality per se, while strong illusionism also denies phenomenality:
I think illusionism about consciousness usually refers to strong illusionism.
Iâm not familiar with the writing of Tye, Dretske, Mike Martin, but what you wrote suggests to me that theyâre weak illusionists and so deny classic qualia, but not strong illusionists, so donât deny phenomenality generally.
FWIW, Iâve seen Michael Graziano, Walter Veit and Heather Browning each self-describe as an illusionist (or something similar) and say they donât like the term and donât like to use it because itâs misleading and confusing.[2] Illusionists are not saying thereâs no such thing as consciousness and are frequently misinterpreted that way, among other ways, like a Cartesian theatre. âConsciousness illusionâ is also probably a confusing term for similar reasons, and something like âillusion of phenomenalityâ would be better.
Iâd also add that being an illusionist doesnât make experiences of red stop seeming to have qualitative features, so it seems to me that some such beliefs are âwired-inâ and instinctual or intuitive, or, as Kammerer (2022) puts it, cognitively impenetrable.[1] You canât get rid of these illusions just by understanding that they are illusions or even how they work, just like you canât for the MĂŒller-Lyer illusion, with which Kammerer (2022) illustrates.
See also Dawson, 2017 for cognitive impenetrability in general, not just in this context.
Graziano, 2016 (ungated) wrote:
It goes on further about resulting misunderstandings the term can cause.
Veit and Browning (2023) (preprint) wrote, responding to some misunderstandings of illusionism:
Apologies if this is derailing the thread by butting in with my thoughts r.e. illusionism. Iâd love to find people to discuss these theories and share thoughts/ânotes outside of this thread. Please get in touch if interested :) maybe we could do a review and adversial/âcollaborative collaboration or something
I will say as a former student of philosophy, and someone who likes reading philosophy a lot more than the median person (though perhaps less than the median EA!) Iâve never been able to get my head around illusionism. Like, I just really donât understand how many people (at least in the rationalist/âEA space) donât seem to grok the Hard Problem. I really think one of the best arguments for âqualia realismâ or the belief that consciousness is a phenomenon demanding of an extra-physical explanation (or perhaps a more convincing one than current theories allow) is a âMooreanâ argument:[1]
I could try and throw in references and arguments but, if Iâm being honest, like every person I do not have the time to re-evaluate each philosophy tradition and argument from scratch, and I find this form of argument very, very strong against the illusionist school, be that Frankish, Dennett, Hofstadter, and illusionist-inclined rationalists (though Iâm actually not sure Yudkowsky is an illusionist here?).
Of course, the illusionists would say that the term is confusing, and that theyâre not eliminativists. Theyâd say that thereâs a difference between consciousness1 - what weâre all experiencing which theyâd agree exists, and consciousness2 - which is the non-physical/âmysterious/âsubjective/âqualitative what-it-is-likeness. This is one of many things, I think, that causes debates on consciousness to often lead to people talking past each other.
Some final thoughts to end (and as I said at the beginning, maybe to discuss on a different place and time):
Iâd recommend trying Sam Harrisâ Waking Up app to explore what introspection tells you. I think itâd be better than many other approaches to meditation which might be a bit ânew-ageyâ for many EA/âLW types. I think the experiences and insight Iâve had with meditation are much, much more convincing to me than what can come across as very confusing, unintuitive, and esoteric arguments in contemporary philosophy of mind.
Sam Harris particularly mentions Douglas Harding, and how Dennett and Hofstadter completely misunderstand his point. Iâm 100% with Harding over Dennett and Hofstadter here. One of Hardingâs students, Richard Lang, has a âHeadless Wayâ course on Waking Up and I think itâs brilliant.
In his great appearance on the 80k podcast, Chalmers brings up an inconsistent triad for the illusionist-inclined to deal with:
Though I suspect that this is again a definitional dispute on what we actually mean by the term âconsciousnessâ
See the conclusion section in this recent Chalmers essay. Kammerer has responded here, but I havenât read that yet
Iâm not sure how much you wanted to get into the object-level here, but Iâll leave a few quick responses to a few points from the (strong) illusionist perspective (or what I understand it to be):
I assume this is supposed to refer to phenomenal consciousness specifically, not consciousness in general, because (strong) illusionists donât deny consciousness in general, and consciousness can be understood in different terms. And, itâs worth noting that people have other illusions that we find hard to disabuse ourselves of on some level, like the MĂŒller-Lyer illusion, with which Kammerer (2022b) illustrates. Itâs intuitively obvious that one line is longer than the other, but itâs also false. The same could be the case for phenomenality (assuming the definition doesnât collapse to one compatible with strong illusionism). Kammerer (2022a) (which you linked to) describes other ways in which we are obviously conscious that are compatible with illusionism: functional and normative.
I would say that this is largely definitional. Consciousness is not entirely an illusion according to strong illusionists; phenomenality (qualitativeness, what-it-is-likeness), classic qualia and dualism are illusions. You can just use the illusionistâs conception of consciousness to ground moral status. This is the approach Dung (2022), Frankish and Muehlhauser take, and a âconservativeâ approach described in Kammerer, 2019. Thatâs also what Iâd do, and Iâd imagine the vast majority of strong illusionists would do.
That being said, I think (stance-independent) moral realism is false anyway (and did so before I became an illusionist), and strong illusionists probably have more reason to be moral antirealists of some kind than most, because similar or even the same debunking arguments would apply to both phenomenality and stance-independent moral claims, e.g. that pain is bad.
Sure, but that only establishes that âitâs intuitively obviousâ is not an infinitely strong reason for belief. It remains a strong one. To overcome the Moorean argument you need to provide arguments for illusionism which are stronger.
Fair. I think the stronger arguments for (strong) illusionism are of the following form:
Physicalism seems true and dualism (including property dualism and epiphenomenalism) false for various reasons.
No other (physicalist) theory besides strong illusionism seems able to address the meta-problem of consciousness or even on the right path.
No theory has an adequate solution to the hard problem of consciousness and some debates between them seem empirically unresolvable (e.g. where the line is between report(ability)/âaccess and phenomenal consciousness), but every theory other than strong illusionism needs to solve it.
There are specific illusionist explanations of some posited phenomenal or classic qualia properties.
There donât seem to be any strong arguments against illusionism (other than possibly mere intuition that phenomenal consciousness is real).
To be clear, Iâm leaving out all of the details, none of the above is obvious, and most or all of it is controversial. I think part of 3 isnât controversial (no full solution yet, and non-illusionist theories need it).
On 4, ineffability and privacy seem easy to explain. First, we donât today know enough of the details of how our brains make the discriminations they do, so we canât fully communicate or compare them in practice yet anyway. Second, even if I understood and could communicate how my brain makes the discriminations it does, this doesnât allow you to put yourself in the same brain states or generally make the same discriminations in the same way. You could potentially build an AI that could or modify your brain accordingly, but this hasnât been possible yet, and it wouldnât really be âyouâ making those discriminations. I canât subject you to my illusions just by explanation, so ineffability is true in practice. With a full enough description, we could compare and privacy wouldnât hold.
Also, I donât take âintuitively obviousâ to be a strong reason for belief, but I am unusually skeptical.
I agree in the sense that other theories canât simply dissolve it, but thatâs almost tautological. If you mean that other theories need to solve it in order to justify belief in them, or in other words if you mean that if we were all certain the hard problem would never be adequately resolved we would be forced to accept illusionism, then I donât think thatâs correct at all.
Consider what we might call âthe hard problem of physicsâ: why this? Why anything? What puts the fire in the equations? Short of some galaxybrained PSR maneuver, which seems more and more dubious by the century, I doubt weâre ever going to get an answer. It is completely inexplicable that anything should exist.
And yet it does. Itâs there, itâs obviously there, everything youâve ever seen or felt or thought bears witness to it, and someone who claims otherwise on the grounds that it doesnât make any sense has entirely misunderstood the nature of their situation.
This is also how I feel about illusionism. Phenomenal experience is the only thing we have direct access to: all arguments, all inferences, all sense data, ultimately cash out in some regularity in the phenomenal content of consciousness. Whatever its ontological status, itâs the epistemic ground of everything else. You canât justify the claim that phenomenal consciousness doesnât exist by pointing to patterns of phenomena, any more than you can demonstrate the nonexistence language in an essay or offer a formal disproof of modus ponens.
So these illusionists explanations are, well, not really explanations of consciousness. Theyâre explanations of a coarse world model in terms of a finer one, but the coarse world model wasnât the thing I wanted explained. On the contrary, it was a tiny provisional step towards an explanation: there are many lawlike regularities in the structure of my experiences, so I hypothesize a common cause and call it âmy brainâ. Itâs a very successful hypothesis, and Iâd like to know whyâgiven that the world is more than just its shadow on the mind[1], why should the structure of one reflect the other?
The illusionist response of âactually your hypothesis is the evidence and your data are just hypothesesâ misses the point entirely.
the dumbest possible solution, but I canât rule it out
The analogy to the âhard problem of physicsâ is interesting, and my stance towards the problem is the same as yours.
However, I donât think the analogy really works.
Is phenomenality itself necessary/âon the causal path here? Illusionists arenât denying consciousness, that it has contents, that thereâs regularity in its contents or that itâs the only thing we have direct access to. Illusionists are just denying the phenomenal nature of consciousness or phenomenal properties. I would instead say, more neutrally:
Note also that the information in or states of a computer (including robots and AIs) also play a similar role for the computer. And, a computer program canât necessarily explain how it does everything it does. âIneffabilityâ for computers, like us, could just be cognitive impenetrability: some responses and contents are just wired in, and their causes are not accessible to (certain levels of) the program. For âusâ, everything goes through our access consciousness.
So, what exactly do you mean by phenomenality, and whatâs the extra explanatory work phenomenality is doing here? What isnât already explained by the discriminations and responses by our brains, non-phenomenal (quasi-phenomenal) states or just generally physics?
If you define phenomenality just by certain physical states, effects or responses, or functionalist or causal abstractions thereof, say, then I think youâd be defining away phenomenality, i.e. âzero qualiaâ according to Frankish (paper, video).
I have no idea what the causal path is, or even whether causation is the right conceptual framework here. But it has no bearing on whether phenomenal experiences exist: theyâre particular things out there in the world (so to speak), not causal roles in a model.
It plays a similar role, for very generous values of âsimilarâ, in the computer qua physical system, sure. And I am perfectly happy to grant that âIâ qua human organism am almost certainly a causally closed physical system like any other. (Or rather, the joint me-environment system is). But thatâs not what Iâm talking about.
Iâm not talking about access consciousness either! Thatâs just one particular sort of mental state in a vast landscape. The existence of the landscapeâas a really existing thing with really existing contents, not a model - is the heart of the mystery.
My whole point is that it doesnât do explanatory work, and expecting it to is a conceptual confusion. The sunâs luminosity does not explain its composition, the fact that looking at it causes retinal damage does not explain its luminosity, the firing of sensory nerves does not explain the damage, and the qualia that constitute âhurting to look atâ do not explain the brain states which cause them.
Phenomenality is raw data: the thing to be explained. Not what I do, not what I say, not the exact microstate of my brain, not even the structural features of my mindâbut the stuff being structured, and the fact there is any.
I donât define phenomenality! I point at it. Itâs that thing, right there, all the time. The stuff in virtue of which all my inferential knowledge is inferential knowledge about something, and not just empty formal structure. The relata which introspective thought relates[1]. The stuff at the bottom of the logical positivistsâ glass. You know, the thing.
And again, I am only pointing at particular examples, not defining or characterizing or even trying to offer a conceptual prototype: qualia need not have anything to do with introspection, linguistic thought, inference, or any other sort of higher cognition. In particular, âseeing my computer screenâ and âbeing aware of seeing my computer screenâ are not the same quale.
But it seems to me that phenomenal aspects themselves arenât the raw data by which we know things. If you accept the causal closure of the physical, non-phenomenal aspects of our discriminations and cognitive responses are already enough to explain how we know things, or the phenomenal aspects just are physical aspects (possibly abstracted to functions or dispositions), which would be consistent with illusionism.
Or, do you mean that knowing itself is not entirely physical?
I think the causal closure of the physical is very, very likely, given the evidence. I do not accept it as axiomatic. But if it turns out that it implies illusionism, i.e. that it implies the evidence does not exist, then it is self-defeating and should be rejected.
I am referring to my phenomenology, not (what I believe to be) the corresponding behavioral dispositions. E.g. so far as I know my visual field can be simultaneously all blue and all dark, but never all blue and all red. We have a clear path towards explaining why that would be true, and vague hints that it might be possible to explain why, given that itâs true, I can think the corresponding thoughts and say the corresponding words. But explaining how I can make that judgement is not an explanation of why I have visual qualia to begin with.
Whether these are also physical in some broader sense of the word, I canât say.
The argument is basically saying that if X canât be explained by physicalism, then X is an illusion. Thatâs treating physicalism as unfalsifable.
No, it isnât just saying that. That understates the case for both physicalism and illusionism that I outlined.
We have good independent reasons to believe physicalism and against alternatives, and I mentioned this, but didnât give examples. Here are some:
Thereâs the good empirical track record of physicalism generally and specifically in giving adequate explanations for the seemingly nonphysical.
There are the questions of where, when, how and why nonphysical properties arise, whether thatâs from or with a collection of particles in a system, over a humanâs development from conception, or in our evolutionary history, that nonphysicalist theories struggle to give sensible answers to. If the nonphysical is fundamental and there at all levels (panpsychism), then we have the combination problem: how does the nonphysical combine to make minds like ours?
Thereâs the expansion of the physical to include whatâs empirically reliable and testable to very high precision and for which we have precise fundamental accounts, including interactions with other fundamental physical properties (although not necessarily all such interactions, e.g. we donât yet have a good theory of quantum gravity). For example, gravity, quantum superposition and quantum entanglement might have seemed unphysical before, but theyâve become part of our physical ontology because of their reliability and our very good (but incomplete) understanding of them and their relationships with other things. Of course, maybe the seemingly nonphysical properties of minds will eventually come to gain the same status, but itâs nowhere close to that now. We shouldnât be hasty to assume the existence of things that donât meet this bar, because the evidence for them is far weaker.
The illusionist also argues (or would want to, but currently lacks the details to make it very convincing) that thereâs a specific adequate (physicalist) explanation for the appearance of X that doesnât require the existence of X. If the appearance of X doesnât depend on its existence, then the appearance of X isnât reliable evidence for its existence. Without any other independent argument for the existence of X (as seems to be the case for phenomenality and classic qualia), then it becomes like any other verified illusion, and our reasons to believe in X become very weak.
Thanks for your response Michael (and your one below to prisonpent). Iâll try to keep it to the point and pre-commit to not responding further as I donât think this is the right place to have a debate about illusionism,[1] but since you presented somewhat of a case for the illusionist I thought I might present the other side.
To me, phenomenal consciousness refers to the first-person perspective, which obviously exists. That first person perspective can make mistakes about the nature of the world, as in the MĂŒller-Lyer case, but I have the experience nonetheless. In the Kammerer(a) piece, he argues that the argument from the anomalousness of phenomenal consciousness is a piece of evidence in favour of illusionism, but I take it one as in favour of non (reductive) physicalism. One manâs modus ponens and all that.
I often find (strong) illusionist writings utterly baffling. I actually re-skimmed Quining Qualia before writing this, and it was really difficult for me to understand[2] even when consulting with GPT-4 in philosopher mode. In Kammerer(a) he refers to a âquasi-phenomenal stateâ, which I have no idea what that is. Again, viewing phenomenal consciousness as the first-person perspective, that just sounds like saying I have âa fake first-person perspectiveâ. To me thatâs the same as saying the first-person perspective doesnât exist, and since it clearly does, there is evidence that illusionist theories do not explain and therefore they are bad theories both philosophically and scientifically.
Iâd be happy to pick this up in an alternate forum though :)
This is partly a me problem, but is also a philosophy problem. Sometimes technical language is needed, but the language of a lot of academic philosophy on all sides often seems to be needlessly obscurantist to me.
Also, the commentary here on Nicholas Humphreyâs views may be illustrative of definitional issues. Humphrey denies the label illusionism for his theory, but Frankish responds that his theory really is illusionist. Also, Schwitzgebel and Nida-RĂŒmelin attempted to define phenomenality as common features of multiple example mental states (and/âor by contrast with unconscious states), but Frankish argues that this doesnât work to define phenomenality (at least not in a way incompatible with illusionism):
Iâm not sure exactly what you mean by âfirst-person perspectiveâ, but strong illusionists might not deny that it exists, if understood in functionalist terms, say.
Frankish says it is like something to be a bat, in terms of a batâs first-order responses or reactive patterns to things, but a bat canât know what itâs like to be a bat, because they donât have (sufficiently sophisticated) introspection on those first-order responses. Dennett says even bacteria have a kind of âuser-illusionâ, because they can discriminate, but only âparticularly reflectiveâ humans are subject to the theoristsâ illusion and worry about things like the hard problem of consciousness. So, we could define first-person perspective in terms of responses or discriminations, and in a way compatible with strong illusionism. This would attribute first-person perspectives extremely widely, e.g. even to bacteria.
If by first-person perspective, you mean introspection, then illusionists wouldnât deny that humans have it.
If by first-person perspective, you mean classic qualia (private, ineffable, intrinsic, etc.), then an illusionist would deny that this exists.
Strong illusionists would also deny phenomenality, of course, in case thatâs different from classic qualia, but some attempted definitions of phenomenality (including what specific physicalist theories define consciousness as, e.g. broadcasting to a global workspace) actually could be understood as defining quasi-phenomenal states, and so compatible with illusionism.
A theory-neutral defintion of quasi-phenomenal states could be that theyâre real things, processes or responses (physical or otherwise) on which introspection (of the right kind) leads to beliefs in phenomenal properties, e.g. these quasi-phenomenal states appear epistemically to us as to be phenomenal. If introspection is reliable and can access phenomenal states, then these accessed phenomenal states would be quasi-phenomenal states under this definition. Illusionists would claim that introspection is not reliable, no phenomenal states actually exist, and so the beliefs in phenomenality are mistaken, hence illusions.
I think it would be wrong to take phenomenal properties as evidence that must be explained, and doing so begs the question against illusionism. What we have evidence of is the appearance of (our beliefs in) phenomenal properties, and illusionism tries to explain that without requiring the actual existence of phenomenal properties. Sometimes (maybe usually) appearances and beliefs are accurate instead of illusions, and the best explanation is based on what they represent actually existing.
I agree. In case of interest: I have published a paper on exactly this question: https://ââlink.springer.com/ââarticle/ââ10.1007/ââs11229-022-03710-1
There, I argue that if illusionism/âeliminativism is true, the question which animals are conscious can be reconstructed as question about particular kinds of non-phenomenal properties of experience. For what itâs worth, Keith Frankish seems to agree with the argument and, Iâd say, Francois Kammerer does agree with the core claim (although we have disagreements about distinct but related issues).
Thanks. (Pleased to see most of this stuff postdates my DPhil and therefore itâs less embarrassing I havenât read it!). I guess I feel I donât really have enough gasp on what phenomenal consciousness is, beyond definition by examples to feel like I entirely understand what is meant by âthereâs consciousness, but not phenomenal consciousnessâ.
I think people generally or often have Nagelâs what-it-is-likeness in mind as the definition of phenomenal consciousness (or at least without classic qualia or nonphysical properties).
If I recall correctly, Frankish (paper, video) called this âdiet qualiaâ and argued that attempts to define phenomenality in more specific terms generally reduce to either classic qualia or âzero qualiaâ (I think purely functionalist terms, compatible with strong illusionism).
Sure, but even the Nagel thing is kind of a metaphor. I find it easy to class which mental states it does or doesnât apply to, but itâs not something I can really characterize in other terms? I donât know, Iâve become less certain I know what all this terminology means the longer Iâve thought about it over the years.
Well, thatâs the whole issue, isnât it? Qualia are the things that canât be fully characterized by their relations.
On some definitions of âqualiaâ yes. I.e. not if you talk in the Tye/âByrne way where âqualiaâ turn out just to be perceived external properties that show up in the phenomenology, for example. And not, necessarily if qualia just means âproperty of a conscious experience that shows up in the phenomenologyâ. But some people do think that about qualia in the second sense, and probably some people do endorse the stronger claim that this is part of the definition of âqualiaâ.
Still having glanced at the Frankish paper I think I get whatâs going on now. Frankish is (I think, didnât read just glanced!) doing something like claiming standard dualist thought experiments show that ordinary people think there is more to consciousness than what goes on physically and functionally, then arguing that this makes that part of the meaning of âphenomenally consciousâ, so if thereâs nothing beyond the physical and the functional, there is no phenomenal consciousness by definition.