The phenomenal concept strategy is a worn-out band-aid

For a quarter of a century, the default physicalist response to arguments for property dualism has been the phenomenal concept strategy (PCS). The strategy turns on a claim about the nature of our phenomenal concepts — the concepts we employ when we attend to our current experiences and think about what they are like. There are many variants of the strategy, and the literature on it is large and technical, but the core idea is simple.

It’s this. Phenomenal concepts function as bare referential devices — demonstratives perhaps. They do not pick out their referents as properties that fit some physical description but latch onto them directly via the exercise of some mental capacity. Since we do not conceptualize phenomenal properties as physical ones, we can easily imagine them varying independently of the physical facts, and this accounts for the intuitions that drive the anti-physicalist arguments — intuitions about zombies, inverts, Mary, and so on. Yet, phenomenal properties might be physical ones all the same. The only dualism the anti-physicalist arguments establish is one of concepts, physical and phenomenal.

I don’t think the strategy works. For it does nothing to explain why we find phenomenal properties anomalous. After all, we employ direct referential devices all the time without generating any ontological worries. Imagine being in a chemistry lab and asking, “What’s this?” or “What sort of stuff is that?”, pointing to a substance or holding up a sample. We don’t conceive of the substance we’re referring to in any particular way — as, let’s say, an acid salt. And, I suppose, we could imagine the substance being removed while all the acid salts remain where they are. But if we are told on good authority that the substance is an acid salt, then we are satisfied. We wouldn’t typically experience any puzzlement as how it could be an acid salt, and we wouldn’t think it conceivable that the stuff we’re indicating could disappear while all the acid salts remained in place. If we were to feel any puzzlement on these matters, it would because of how the stuff looked and our background beliefs about what acid salts are like.

If the PCS were sound, then the same should be true of phenomenal properties. Gesturing inwards at a twinge of pain and asking “What’s this?”, should not generate any intuitions about the nature of the state picked out and we should have no difficulty in accepting that it is a physical one, if that’s what the science indicated.

But that doesn’t happen. Even if we’re thoroughly convinced that the brain has no nonphysical properties, we still feel puzzled by the situation. We still can’t understand how this twinge of pain could be a brain state, and we still have a strong inclination to think that there’s some extra feature present that is only contingently connected to the physical.

The moral I draw from that is that phenomenal concepts are not bare referential devices. They incorporate some substantive conception of their referents. If they are demonstratives, they have a tacit theoretical sortal attached. We wonder, not simply, “What’s this?”, but “What’s this phenomenal feel?”.

What is this substantive conception of the phenomenal? I think it’s roughly the one Daniel Dennett dismantled in “Quining qualia” — the concept of a qualitative state that is private, ineffable, intrinsic, and immediately apprehended. Maybe those commitments are qualified in various ways, but they are still strong enough to make the conception incompatible with our conception of the physical. Hence our puzzlement.

If that’s right, then there’s only one option for the physicalist, and that is to say that phenomenal concepts misrepresent their referents. The properties they pick out (assuming they pick out determinate properties at all, which they may not) aren’t really phenomenal ones. And that’s illusionism.

For twenty-odd years, the PCS has acted as a band-aid holding physicalism and phenomenal realism together, and it’s worn out.

Accelerating research on consciousness?

Back in December, psychologist and author Christian Jarrett got in touch to ask what I thought about the new project “Accelerating Research on Consciousness” organised by the Templeton World Charity Foundation. See this news story for more information about the project. Christian incorporated some of my comments into an article for BBC Focus magazine (which I recommend) but I thought I’d post my full reply here, in case anyone is interested. Here it is.

I have mixed feelings about the project. I’m delighted to see more funding for experimental work on consciousness. The data collected will undoubtedly be useful. I have worries, however. It looks like the project will focus on explaining consciousness in the phenomenal sense. That is, the organizers and participants conceive of conscious states as essentially subjective ones, involving awareness of phenomenal properties or qualia (the private mental ‘feel’ or ‘what-it-likeness’ of experience). If that’s right, then I am dubious of the chances of making decisive progress.

To begin with, it’s hard to see how one could explain phenomenal properties in terms of brain processes. The two things are just too different. (This is the so-called ‘hard problem’ of consciousness.) The most we can hope to do is to find correlations between brain processes and phenomenal properties. And even then there’s a methodological problem. For there can be no objective test for the presence of essentially subjective properties. The best we can do is to test for objective indications of their presence, such as the subject’s reports and reactions. And this means that tests of correlation hypotheses can never be decisive. Suppose theory A says that conscious state C occurs when brain region N1 is active, whereas theory B says that N1 isn’t sufficient on its own and that brain region N2 needs to be active as well. And suppose we run some experiments and find that participants report C when both N1 and N2 are active but not when only N1 is. Does this prove that theory A is wrong and theory B right? No. It might be that N1 is sufficient for C, but that N2 is needed to enable us to report it. The same problem will arise if we try to test for nonverbal indications of C. Again, how do we tell which brain states are necessary for the conscious state itself and which are necessary for producing the behavioural indications of it? Since there is no way of directly testing for subjective properties, we can never definitively rule out any theory.

In short, so long as we focus on phenomenal consciousness, we’re never going to have decisive tests of our theories. The moral I draw is that shouldn’t focus on phenomenal consciousness. In fact, I believe that we do not have phenomenal consciousness; it’s a kind of introspective illusion, which reflects the limited access we have to our own mental processes. (I call this view ‘illusionism.’) The real task is to explain our intuitions about phenomenal consciousness — why we think we possess it.

As regards the theories currently being tested, I am very sceptical of IIT. It is intended as a theory of phenomenal consciousness, so the worries I’ve just mentioned apply, but even as theory of that, IIT is implausible. All kinds of things can have a rich informational structure in the relevant sense, so the theory has the consequence that inanimate objects can be phenomenally conscious. Even a blank wall could be.

I am much more sympathetic to Global Workspace theory, though I think it should be construed as a theory of access consciousness — of the awareness of information in a functional sense — rather than phenomenal consciousness. Moreover, it needs to be supplemented with some account of why we think we have phenomenal consciousness.

As for what I’d like to see next: Unsurprisingly, I’d like to see the project test illusionist theories of consciousness, which focus on explaining our intuitions about phenomenal consciousness. These do not face the problems I’ve mentioned, and they offer a promising line of research. It’s early days yet, but such theories are being developed. A good example is the Action Schema Theory proposed by the Princeton neuroscientist Michael Graziano and his colleagues.

The bottom line, then, is that the funding for experimental work is welcome and the data gathered will be useful, but the project is unlikely to settle anything until we have a better conception of exactly what it is we are trying to explain.

Bright shiny colours

What are colours? My view is that they are properties of surfaces in the world around us — albeit complex gerrymanded ones, which can be picked out only by reference to our reactions to them. Blue things are things that evoke a certain distinctive cluster of reactive dispositions in us. Note that that I do not say that they are ones that produce blue sensations in us. I don’t think that experiencing blue involves entertaining a mental version of blueness — a blue quale or phenomenal property.

Where then is the quality of blueness ? It’s not out there in the world. Out there there’s just a surface with a microstructure that reflects certain wavelengths of electromagnetic radiation. And I’ve denied that there is any blue quality in our minds. So where is the blueness of the blue?

My answer is that it is not really anywhere. It’s a property that our minds misrepresent external objects as having. However, it’s a property that corresponds to, and carries information about, something real and important — namely, the affordances of the objects in question. That needs a lot of unpacking and qualification, but the general idea is this. We are tuned up, by biological evolution, cultural evolution, and personal experience, to track worldly properties that it’s useful for us to notice. Such properties afford us opportunities for action in various ways; they have specific affordances. An object’s affordances are reflected in the suite of reactive dispositions its perception triggers in us — the suite of beliefs, expectations, associations, emotions, priming effects, and so on.

Now my suggestion is that the human brain monitors its own reactive dispositions and generates schematic representations of them, which are linked to its representations of the objects that triggered them. The upshot of this is that we experience the world as being metaphorically coloured by our reactions to it. We experience objects as having a distinctive but ineffable significance for us, which is a marker of their affordances. This is what we call their quality or feel. The blueness of blue is a distorted representation of the affordances it presents, represented as a property of the object itself.

That’s still very schematic, but a little example may help. Consider shiny, metallic colours, such as silver and gold. These seem to have a distinctive feel to them, and as a child I was very puzzled as to where they fitted into the visible spectrum. But, of course, they are not really different colours. Shiny things are just regularly coloured things whose brightness (and colour if they are very shiny) varies markedly with viewing angle. What gives them their distinctive ‘feel’ is precisely the affordances they present. We expect them to change in a distinctive way as we move in relation to them. The ‘feel’ of metallic colour just is the expectation of this effect.

A postscript: Another illustration of this is afforded by Gregory Thielker’s paintings of scenes though rain-spattered glass. In me, these create a powerful response (‘feel’, if you like). Doubtless, this is in part because they evoke memories of glum hours spent in traffic during rainy commutes. But I think it also reflects the way they trigger strong expectations that the scene will morph and distort in a distinctive way as the water drips or I move my head.

Illusion or identity?

Illusionists believe that consciousness involves no properties that are not detectable and fully describable by third-person science. Any other properties we think are involved are illusory. Suppose that’s right. Still, why should it follow that phenomenal properties are illusory? Why not say that they are properties that are detectable and fully describable by third-person science? It’s true (the objection continues) that we think of phenomenal properties as ones that present a problem for science — that pose a hard problem — but it doesn’t follow that they really do present one. Maybe we are just wrong about them.

Suppose that phenomenal concepts do in fact track completely unmysterious brain properties, which for some reason we mistakenly think of as nonphysical. There are many candidate explanations of why we might do this. If that’s the case (and illusionists don’t deny the possibility), then wouldn’t it be better to say that phenomenal properties are real but different from what we thought?

Here’s my answer. Maybe we could say that. It’s a revise-or-eliminate situation, and there is no simple procedure for determining the best way to go. But here are some reasons for rejecting the revisionary route.

First, it would invite confusion. The concept of the phenomenal carries a lot of connotations that physicalists must reject — assumptions about the reliability of introspection, intuitions about well-known thought experiments, associations with dualist notions such as sense data, and so on. Using a term with all this theoretical baggage is not the most perspicuous way of presenting a physicalist theory of consciousness.

Second, it would be misleading. The notion of phenomenal consciousness has become bound up with that of the hard problem — a problem that is supposed to be both substantive (there’s a real thing that needs explaining) and qualitatively different from ‘easy’ problems that can be solved by cognitive science. To offer a theory of phenomenal consciousness is to suggest that one has solved this hard problem, and physicalists shouldn’t do that. For physicalists, there is no hard problem, only the problem of explaining why there seems to be one.

Third, it would be tedious. In theoretical work, we’d have to laboriously disinfect phenomenal concepts before use, explicitly disavowing all their theoretical accretions.

Fourth, it would be pointless. After disinfection, we’d be left with nothing more than a bare demonstrative or quotational device, equivalent to ‘whatever this is’, applied introspectively. It’s not clear that this would pick out something determinate or theoretically interesting. We’d be gesturing at the whole complex perceptual-cum-reactive state triggered by the current stimulus, and without further specification it’s doubtful that the gesture would pick out a clear target for scientific investigation. (By contrast, gesturing at the supposed qualitative aspect of the state would narrow down the target, but only to something that physicalists must say is illusory.)

Fifth, it’s restricting. Physicalists need phenomenal concepts in their old theoretically laden senses in order to describe how people mistakenly think of consciousness (‘It seems that experiences have a phenomenal aspect as well as a functional one’). Compare the term ‘witch’. If we revise it to mean female naturopath, then it becomes harder to express what mediaeval people thought. After all, they were right to think that there were witches in that sense. Of course, this is only a linguistic problem and it could be solved by paraphrase, but it’s a consideration.

In the end, the concept of the phenomenal is too compromised to be useful to science. As Daniel Dennett says in his Consciousness Explained, let’s cut the tangled kite string and start over. Phenomenal properties are illusory.

The trolley problem murder

Moriarty wants to kill Holmes but doesn’t want to run the risk of being convicted for murder. So he plans to get an innocent person to do the dirty work for him. Here’s what he does.

First, he persuades Holmes (with whom he is ostensibly on good terms) and three of his Baker Street Irregulars to take part in a real-life philosophical experiment. He will tie the three Irregulars to a trolley track, near to a spur on which Holmes will be tied. He will wait till there are a number of bystanders near where the spur branches off, then release a trolley down the main track in the direction of the Irregulars, and wait to see if one of the bystanders throws the lever to divert the trolley onto the spur. The experiment will provide the philosophical world with important data on the famous trolley problem.

Moriarty shows everyone that the trolley has an automatic braking mechanism that will prevent it hitting the Irregulars if the lever is not thrown. This mechanism really does work, but Moriarty has adjusted it so that if the trolley is sharply diverted onto another track, it will fail. Everyone agrees to take part.

It all goes to plan. Moriarty releases the trolley, a bystander, Doyle, sees the trolley and the people on the tracks, does some rapid moral reasoning, and operates the lever, believing that this will result in the death of the person on the spur. The trolley goes onto the spur, its braking mechanism fails, and Holmes is killed.

Moriarty’s plan has worked. Holmes is dead, yet Moriarty did not kill him. If Doyle had not intervened, no one would have been harmed. Instead, Doyle freely chose to do something which he believed would kill Holmes.

Or was there a flaw in the plan? Did Moriarty commit a crime? If so, what was it? Who is morally responsible for Holmes’s death? Does it matter if you think Doyle’s moral reasoning was faulty?

(Revised and expanded after twitter discussion with @JulianSales2, @FillinghamLydia, and @tylerhower.)

The evil trolley problem

Photo of child on railway line

On Tuesday I posted a question on Twitter (prompted by a remark by @bowmanthebard). The question involved a version of the trolley problem, in which one has to choose between letting die and actively killing, but with the twist that the motive for choosing the active option is not to minimize loss of life (in this version it might actually increase it) but to preserve a life especially dear to one.

As the question generated some interest, I thought I would set out the thought experiment in more detail here. Another Twitter user @gjfitzgerald described my question as evil, since there is no way to answer it without guilt, so I call it The Evil Trolley Problem (not a perfect name, I admit, since it’s ambiguous). I’m not an ethicist, and I don’t know if this particular scenario has been previously discussed in the literature (if it has, I’d be grateful for references), though I’m sure the underlying issues have.

Here is the problem:

You are walking by the railway line, where a group of local children are playing. Suddenly, there is a shout and you see that a runway trolley is thundering down the track. You run to warn the children and see that one child is trapped on the line. With horror you realize that it is your own child. The trolley will certainly kill your child if you do not act. Luckily, you are close to the points, and by operating a manual lever you can divert the trolley onto another track. But as you grasp the lever, you notice that another child, not known to you, is trapped on the other track. If you pull the lever, the trolley will certainly kill them.

What would you do? There are only seconds left and there is no other option. Would you sacrifice an unknown child to save your own? If you would switch the trolley, would do the same if it it would result in more deaths? What if there were two children trapped on the other line or if there were a school bus stalled there? How many children would you sacrifice to save your own child?

The scenario is of interest because what most of us would do in the imagined situation is in contrast to what moral theory tells us we should do. I suspect that most of us would switch the trolley, even if it would result in many deaths. Yet I doubt if there are many moral theories that would dictate that course, or even judge it permissible, and most legal systems would, I assume, class it as murder. This in turn raises wider questions about how far moral theory should bend to human nature, and how we can reconcile our intense preference for our kin with our ideals of altruism and egalitarianism.

Image credit

Philosophy article recipe

1. Take one everyday concept and analyse it carefully. (Vegans may substitute an x-phi survey.)

2. Stir in two or three strong intuitions and a cupful of raw scientific data. Beat the mixture vigorously until it looks like a hard problem.

3. Add half a cup of finely chopped logic, two heaped tablespoons of folk psychology, and some freshly prepared terminology.

4. Spread the mixture thinly over a standard article format and glaze with superficially novel proposals.

5. Allow to rise over a conference season.

6. Place into a high-impact journal and leave for 6 to 24 months. Test occasionally with an email. The article is ready when it turns to a crisp PDF.

7. Garnish with impact factor and present in a CV or funding application.

Serves: None.

Cartoon of chef

Something that it is like to be

‘[F]undamentally an organism has conscious mental states if and only if there is something that it is like to be that organism — something it is like for the organism.’

So writes Thomas Nagel in his famous 1974 paper ‘What is it like to be a bat?’ (p.436). It is a compelling thought and one that seems self-evidently true. (I remember coming up with the same idea when I was a teenager and thinking it a great insight, and I’m sure many others have had the same experience.) But is Nagel’s claim correct or might it be seductively wrong? It depends, I think, on how we interpret it. Here are two things it might mean.

The first is that having conscious experiences involves having a certain kind of introspective self-awareness — an awareness of one’s own mental responses to the world. For you to be conscious is for you to know what you are like as you respond to the world, as well as what the world is like as it affects you. I’ll call this introspective subjectivity. Introspective subjectivity, we might say, stresses the ‘like’ part of ‘what it is like to be’.

The second thing Nagel might mean is that to have conscious experiences is to have a form of immediate inner awareness that exists simply in virtue of being the thing one is and that is not dependent upon introspective mechanisms. I’ll call this intrinsic subjectivity. We might say that this stresses the ‘be’ part of ‘what it is like to be’.

These two kinds of subjectivity are very different. Introspective subjectivity is not essentially private. With the right apparatus, another person might monitor the same internal states my introspective mechanisms do and so share my introspective awareness. But intrinsic subjectivity cannot be shared. The only way someone could share what it’s like to be me in the intrinsic sense is to be me, or perhaps a duplicate of me.

Now there doesn’t seem to be any special difficulty in understanding how a creature could possess introspective subjectivity. it would just need to have suitable introspective mechanisms targeting its own internal responses and hooked up in the right way to the rest of its cognitive system. But intrinsic subjectivity looks like a complete mystery. How does this inner awareness arise? What exactly is the subject of it? Which things have it? What does it do? How can we even investigate it? These look like, well, hard problems.

In the literature on consciousness these two kinds of subjectivity aren’t always clearly distinguished. Some theories — most obviously panpsychist ones — are plainly theories of intrinsic subjectivity. If there is something it is like to be a rock, it’s not because the rock is capable of introspection. But other theories look like theories of introspective subjectivity. Higher-order representational theories, for example, attempt to explain consciousness in terms of the internal monitoring of experience. Yet these theories are often discussed as if they were alternative accounts of the same thing.

Nagel’s claim has become the standard starting point for theories of consciousness, but it doesn’t identify a unique explanandum and it has sent researchers off down very different paths. In my view, it is immensely plausible to think that conscious experience involves introspective subjectivity, and developing theories of introspective subjectivity should be a major research programme for cognitive science. But the pursuit of theories of intrinsic subjectivity is, I fear, misguided and futile.

Image credit

Is the hard problem an illusion?

An oddly shaped iceberg

Note: This is a revised version of a letter I sent to The Guardian, responding to a letter by Philip Goff, which itself commented on an article on consciousness by Oliver Burkeman. The letter was deemed too long for publication in the paper, so I am posting it here instead. It is written for a general audience.

As a member of the Daniel Dennett camp on the Greenland consciousness cruise referred to in Oliver Burkeman’s article, I should like to respond to Philip Goff’s letter of 28 January 2015. Goff advocates a radical solution to the Hard Problem of explaining how consciousness fits into the natural world. Consciousness, he argues, is not a physical process, but an intrinsic feature of all physical reality. Consciousness is not fundamentally material; rather, matter is fundamentally conscious. A consequence of this view is that everything is conscious to some degree: trees, stones, atoms, quarks — all have a little bit of consciousness. This panpsychist position offers a neat solution to the problem, and Goff argues for it with intelligence and elegance, but I find it hard to take it seriously.

I do agree with Goff on one important point: Consciousness, as we ordinarily conceive of it, cannot be explained by the physical sciences. The Hard Problem, as posed by David Chalmers, can’t be solved by cognitive science. Goff draws the moral that consciousness is not physical in the ordinary sense. I draw the moral that we are conceiving of consciousness wrongly. We are mistaken about what consciousness is.

Our conception of consciousness is derived from introspection — from mentally ‘looking inwards’ at our experiences. When we do this, our experiences seem to have a private ‘phenomenal quality’ to them (think of the sensation of seeing a vibrant green leaf, or smelling coffee grounds, or running one’s fingers over a silk scarf). These phenomenal qualities (or ‘qualia’) seem almost magical and utterly different from the mundane physical properties of our brains.

But maybe that’s an illusion. Maybe when we introspect, what we are aware of are certain patterns of brain activity that seem magical and nonphysical but aren’t really. Moreover, as another cruise participant, Nicholas Humphrey, argued, maybe these brain processes were shaped by evolution precisely to seem magical to introspection. In his 2011 book Soul Dust Humphrey argued that evolution adapted pre-existing neural systems to create an inner ‘magic show’ which carries immense adaptive benefits — enriching our lives and our experience of the world, enhancing our sense of self, and deepening our engagement with each other. In short, maybe evolution has hardwired us to think that we have a magical inner life, and the problem of consciousness is a benign trick that nature has played on us.

Most people, I find, think this suggestion is just as crazy as panpsychism. If there’s one thing we are absolutely certain of (the argument goes) it’s our experience. We may doubt that there is a green patch in front of us, but we can’t doubt that we are having an experience with a green phenomenal quality. This takes us back to the origins of the Hard Problem in Descartes’ sceptical thought experiment mentioned in Oliver Burkeman’s article. There’s something right about this. If we suspect that our senses are misleading us about the external world, then we retreat to more cautious and secure claims about how things seem to us. But (I would argue) such claims should not be construed as infallible reports of the nature of our experiences. Being cautious about the external world doesn’t make us infallible about the interior one. We may be sure that we’re introspecting something, but can we rule out the possibility that we’re mistaken about its nature, just as we may be about the nature of external things? After all, to the spectator a good illusion of something is indistinguishable from the thing itself.

Of course, it’s not so simple to solve the problem of consciousness. For one thing, we need to explain what it means to say that experiences seem to have phenomenal qualities. (It better not mean that they generate further experiences which really do have phenomenal qualities. Otherwise we’d merely have moved the Hard Problem back a step.) But thinking of consciousness as involving an illusion changes the questions we have to answer, and does so, I believe, in a productive way.

On the cruise I proposed the name ‘illusionism’ for the sort of position I have been describing, and the term ‘the Illusion Problem’ for the problem of explaining how the consciousness illusion is created. (I wasn’t claiming to have originated the position or the problem; Daniel Dennett has advocated illusionism for decades, and Nicholas Humphrey has done pioneering work on the Illusion Problem.) For me, the attraction of illusionism is that it allows us to give full weight to the intuitions that motivate views like Goff’s — consciousness really does seem weird — without requiring us to endorse a weird metaphysics. Maybe it’s time to stop banging our heads against an illusory Hard Problem and start trying to solve the hard-ish but solvable Illusion Problem?


The talks from the consciousness cruise, including Jesse Prinz’s introduction to my paper on illusionism, my reply, and the following discussion, were videoed by the Moscow Center for Consciousness Studies and can be viewed on the centre’s Youtube channel. Here is the full playlist.