A bold draft paper from Tom Clark tackles the explanatory gap between experience and the world as described by physics. He says it’s all about the representational relation.
Clark’s approach is sensible in tone and intention. He rejects scepticism about the issue and accepts that there is a problem about the ‘what it is like’ of experience, while also seeking to avoid dualism or more radical metaphysical theories. He asserts that experience is not to be found anywhere in the material world, nor identified in any simple way with any physical item; it is therefore outside the account given by physics. This should not worry us, though, any more than it worries us that numbers or abstract concepts cannot be located in space; experience is real, but it is a representational reality that exists only for the conscious subjects having it.
The case is set out clearly and has an undeniable appeal. The main underlying problem, I would say, is that it skates quite lightly past some really tough philosophical questions about what representation really is and how it works, and the ontological status of experience and representations. We’re left with the impression that these are solvable enough when we get round to them, though in fact a vast amount of unavailing philosophical labour has gone into these areas over the years. If you wanted to be unkind, you could say that Clark defers some important issues about consciousness to metaphysics the way earlier generations might have deferred them to theology. On the other hand, isn’t that exactly where they should be deferred to?
I don’t by any means suggest that these deferred problems, if that’s a fair way to describe them, make Clark’s position untenable – I find it quite congenial in general – but they possibly leave him with a couple of vulnerable spots. First, he doesn’t want to be a dualist, but he seems in danger of being backed into it. He says that experience is not to be located in the physical world – so where is it, if not in another world? We can resolutely deny that there is a second world, but there has to be in some sense some domain or mode in which experience exists or subsists; why can’t we call that a world? To me, if I’m honest, the topic of dualism versus monism is tired and less significant than we might think, but there is surely some ontological mystery here, and in a vigorous common room fight I reckon Clark would find himself being accused of Platonism (or perhaps congratulated for it) or something similar.
The second vulnerability is whether representation can really play the role Clark wants it to. His version of experience is outside physics, and so in itself it plays no part in the physical world. Yet representations of things in my mind certainly influence my physical behaviour, don’t they? The keyboard before me is represented in my mind and that representation affects where my fingers are going next. But it can’t be the pure experience that does that, because it is outside the physical world. We might get round this if we specify two kinds of representational content, but if we do that the hard causal kind of representation starts to look like the real one and it may become debatable in what sense we can still plausibly or illuminatingly say that experience itself is representational.
Clark makes some interesting remarks in this connection, suggesting there is a sort of ascent going on, whereby we start with folk-physical descriptions rooted in intersubjective consensus, quite close in some sense to the inaccessibly private experience, and then move by stages towards the scientific account by gradually adopting more objective terms. I’m not quite sure exactly how this works, but it seems an intriguing perspective that might provide a path towards some useful insight.
I haven’t read Tom’s paper yet, but I dissent from the setup in the paper’s abstract. The key point is that we should use a promiscuous interpretation of “physical property”. Analogous to unrestricted mereological composition, we should allow that any arbitrary hypernym of names of physical properties, names another physical property. So for example, “reflects only photons with wavelengths 620 to 750 nm OR reflects 610 nm and 750 nm photons in equal numbers” is a physical property. By stringing enough OR clauses together, we can get at objective redness.
From a particle physicist’s viewpoint, that property may seem bizarre and useless. But in psychology or lighting design, it’s a highly useful property. That we carve this property out with an anthropocentric viewpoint is neither here nor there, and should not undermine our judgment that this is a physical property. Nor should the fact that there are some fuzzy boundary cases. Many of our most central physics concepts, from “pressure” to “granite” to “causality”, admit of fuzzy boundary cases.
Is subjective redness a physical property? There are an awful lot of physical properties, so to rule it out, we would need a very general argument.
Peter, many thanks for both your sympathetic and cautionary remarks, the latter well taken. I heard the ice cracking as I skated by the many complexities concerning representation and content that are being addressed by lots of smart people these days, including Andy Clark, Anil Seth, Jesse Prinz, and Thomas Metzinger to name just a few. I’m pinning my hopes on some sort of theoretical convergence on consciousness as the science of representation develops, we shall see.
As you note, I’m not advocating any sort of substance dualism, but the two sides of the representational relation – representational reality vs. represented reality – might be construed that way. There’s clearly a major role for physically-realized intentional content in explaining behavior (e.g., as in the beliefs and desires of Dennett’s intentional stance), but I don’t see *phenomenal* content per se (conscious experience) as being in a position to participate in 3rd person, physically-based explanations, only in 1st person explanations (those that advert to conscious experiences as causal, see section 7 on phenomenal-physical explanatory parallelism). The physical level of the representational vehicles (e.g., neural networks) will always be intuitively construed as the “hard, causal” hence really real level of explanation, but neither intentional content nor phenomenal content and their respective explanatory spaces are going away. And of course the physical level is what’s *represented to be the case*, whether in terms of sensory experience and/or scientific theory. So it’s not as if it has ontological priority over the representations in terms of which it’s presented to us knowers, or so it seems to me.
Lots to chew on here, obviously, so I look forward to any comments from the usual suspects. Thanks again!
“He says that experience is not to be located in the physical world – so where is it, if not in another world? We can resolutely deny that there is a second world, but there has to be in some sense some domain or mode in which experience exists or subsists; why can’t we call that a world?‘
The problem with calling the Platonic realm a world (or a realm) is that we delineate and navigate our physical world by physical measures and we risk using similar measures when they don’t/can’t apply. Similarly when we try to use terms like “exist”. So does the number “three” exist? Did it exists before people? Before life?
I solve this problem for myself by having my ontology be made of two things: stuff and patterns. I can then say stuff exists, and patterns are real. Some patterns are directly discernible in stuff, such as the “three” in three apples. All patterns are, potentially, indirectly discernible in stuff as symbolic signs, such as the words “three apples”.
So I think Clark is on the right track in that consciousness is a pattern, specifically, a pattern discernible in stuff. For human consciousness, the pattern involves activities of neurons which amount to symbolic signs, such as a sign which represents “three apples”. The “three apples” is a concept, which is a pattern, so is not a physical thing. But if we want to refer to that pattern, as a past event for example, the only handle we have on the event is the reference of the symbolic sign. We can’t reference the sign vehicle.
*
see what you make of Ray B @ https://iai.tv/video/matter-and-mind
“key moments” pitch @ 8:30 mins
It’s possible I’m missing something, but I’ve never really seen the philosophical problem with representations. It seems like there are multiple plausible theories about what’s going on. My current preferred one is that representations / image maps / mental concepts are predictions, or more accurately, prediction frameworks. But even if that eventually turns out to be wrong, this doesn’t strike me as an intractable problem.
On Platonism, James’ remarks largely capture my own view. But I would add that if we start thinking of the patterns as something separate from the physics, it pays to remember that a lot of stuff we consider physical: molecules, atoms, protons, neutrons, are composite patterns of more primal stuff. And it may be that what we consider to be elemental particles may themselves eventually turn out to have constituents. Physics may be structure all the way down.
Along those lines, I think Tom is right that experiences exist. (I prefer saying they’re constructed rather than illusions.)
But I’m uneasy with the assertion that they’re not physical. I see the relation between the experienced representation and the neural firing pattern similar to the relation between a software bit and a transistor. It might appear that the bit and other software entities have a separate Platonic existence, but we know that the bit maps directly to a transistor state, and that software concepts all map to processor states. (A bit once mapped directly to a vacuum tube and may map to something else in the future.) The mapping between a neural firing pattern and a particular quale are far more complicated and definitely not understood yet, but it seems like cognitive neuroscience is always making progress on that front.
Of course, people will argue that these will only amount to correlations, and they’re right. But correlations are all we ever get. We can narrow them down, hopefully to single variables, but we can never get more than that.
SelfAwarePatterns, you say you are uneasy with the assertion that experiences are not physical. You say the relation of an experienced representation to a neural firing pattern is similar to the relation of a bit to a transistor. I think you have mistaken the relationship. The relation of a bit to a transistor is more like the relation of specific firing rates of individual neurons to the overall pattern. The relation described is a description of the relevant physical readout out of the system.
So, for example, you could say that the bit readout of a transistor at a given time is 1. But then you could say that the readout of a set of neurons (just 3, for simplicity) might be 10, 0, and 20 hz, respectively.
My point, if not Tom’s, is that the experience isn’t the readout, it’s what the readout is for. So the experience might be the concept “red ball”, if that’s what that pattern (10,0,20 hz) means to the system. That one bit from the transistor could just as easily mean “red ball” if the system is set up right.
Does that make sense? Does it make any difference?
*
Through physics, we’re slowly beginning to understand we represent experiences…
…Hence, we are, as representational experiences, also as, ‘fundamental forces-interactions of the universe’…
Having been given place in the universe, what’s next, ‘paths towards some useful insights’ sounds good to me…
…”towards” could be drafted-posited to be the ‘object’ of ones own insightful experience…
James #6,
I think you might be confusing where a bit falls between hardware and software. A bit is a software concept. It’s value of 1 or 0, or true or false, exists at a certain level of abstraction, I think the lowest level of abstraction in software.
But the equivalent of the firing rate of neurons, a very mechanistic description, for a transistor, might be the current voltage level of the transistor. One range of voltage states maps to a bit equal to 1 or true, and another range to 0 or false.
The “red ball” concept isn’t quite a good match for the bit, because the ball is much more complex. In software or in minds, we might imagine the ball concept being signaled, maybe a specialization of the spherical concept, and perhaps a redness concept. In software, this could conceivably be a collection of bits, each a flag indicating certain attributes of the object, which in turn map to transistors. In a brain, it might be various qualia that map to signals from various synapses, one for redness, one for ballness, etc.
Hopefully that clarifies rather than obscures.
Now that I read the paper, I think it’s awesome. Which doesn’t mean I’m convinced, but there are some real gems. “[T]he maximally objective, quantified specification of the world – the spatio-temporal world as described by science – is still a representationally encoded model, not the unrepresented world itself.” Yup. And I learned that since grad school, some phil mind theses I’ve long favored have acquired a name: enactivism.
Now let’s turn to the privacy argument. Suppose I have a private tunnel to the Sultan’s Palace. I’ve used it many times, and never encountered anyone else in that tunnel. Is it *exclusively* private, i.e. proprietary? I can’t know; all I know is, at most, that nobody else has used it *yet*. Nor do I have the *perception* (literally) that it is exclusive, although if I’m a hasty generalizer, I might have that *thought*.
Thus, the claim that pain only exists for you, the experiencer, begs the question against Churchland, Tye, Prinz, et. al. Just because an identity claim isn’t obvious doesn’t mean it’s not true. And if Clark Kent keeps on showing up everywhere that Superman does, eventually smart observers should suspect that Clark Kent *is* Superman. Clark Kent standing for (e.g.) neurally-instantiated attended intermediate representations, in case the metaphor isn’t clear.
SelfAware #8,
Your comment clarifies how you are missing my point. 🙂 The kind of representation I think Tom is talking about, and I’m talking about in any case, is where the physical vehicle for the sign is essentially arbitrary, which allows even a simple binary sign to convey complex information if established in advance. “One if by land, two if by sea.” Again, the point is that the thing being referenced, the object of the reference, is a non-physical thing. It’s a pattern. (This works for the bit when that particular bit stands for “true”.)
Now in the brain I think it is unlikely that one conce
[ahem] [wish there was a way to edit mistakes]
Now in the brain I think it is unlikely that concepts are associated with individual synapses. Chris Eliasmith has demonstrated biologically plausible mechanisms for representing concepts such that they can be combined and uncombined when needed. He calls them Semantic Pointers. So the semantic pointer for “red ball” might be a hundred or so neurons firing in a particular pattern. When those neurons are firing in that pattern, and the subject (human or computer) is responding to (attending to) that pattern, that subject is experiencing that pattern: “red ball”.
I’m sure it’s all clear now.
*
[he said]
…consciousness as independent in our universe is helped by learning to see oneself as a object representing experience…
My objectifiable objection…
James,
I actually think we’re on the same page here. My point is that each layer of abstraction defines what the lower layer means. So a bit defines what a particular transistor state at a particular instant means, a pixel on a screen may define what that bit means, the overall pattern of the pixels, the image, defines what that pixel means, etc. Each layer is its own causal framework. We can talk about the operation of computers in terms of the high level software logic, the lower level logic, or purely in terms of how transistors shape the pattern of electrical flow. I think brains are much the same, although evolution has no obligation to have its layers have a clean engineered like consistency.
I can understand the reluctance to see individual synapses as being meaningful at too high a level. My sense is that it would be more like the bit layer in computing than the ball layer. On the other hand, Google “Jennifer Aniston neuron”. Of course, there’s nothing magical about that neuron that in and of itself makes it about Jennifer Aniston, just that it’s the final tip of an activation hierarchy that ends with it.
Paul, thanks for the kind words and it’s good to know that you think the paper might even be wrong. Re privacy, you say “…the claim that pain only exists for you, the experiencer, begs the question against Churchland, Tye, Prinz, et. al. Just because an identity claim isn’t obvious doesn’t mean it’s not true.”
Seems to me that it’s necessarily the case that only I can experience my pain, since pains and other experiences are a function (somehow) of being a discrete system, so are not shareable. They aren’t public objects potentially available to observation, unlike the system itself. True, we might be able observe all the physical correlates of a pain down to the last detail, but the pain itself is somehow a function of instantiating those correlates. It exists only for the instantiating system. This seems to me not to beg any question, but rather what needs explaining.
Colors of course are more easily supposed to be objective physical properties since they are ordinarily (but not always) experienced in conjunction with external physical objects (hence they are widely used in defenses of physicalism). But as I think Dennett’s afterimage example ironically ends up demonstrating (section 3 of the paper), they are bits of (real) phenomenal content that we then commonsensically attribute to objects as their properties. But as content items they aren’t locatable, unlike the objects that appear in terms of them. Likewise, pains are represented properties of the painful limb, tooth, or whatever, but when we go looking for pain in the system as an objective observable physical property, we won’t find it, only its physical and functional correlates (the point of Dennett’s “Why you can’t make a computer that feels pain”).
What if looking seeing consciousness were not ever related to observation…
…if the intention to survive sense feel think be, are simply possessions of living systems…
Is intention-attention different than observation…
…that evolution is an object a observation…
Tom, I would say – maybe tendentiously but that’s OK in this context – that conceivably, I can visually experience (see) your pain on the screen of an fMRI machine, if I’m a neurologist who knows an enormous amount about human pain. So you should reword your claim to “necessarily, only I can *be hurt* by my pain.”
But even that’s too strong. Take Parfit’s thought-experiment about persons who can fission like amoebas and also fuse together into one. You and I start separate, fuse together for a day, then fission apart again into physically and psychologically recognizable Tom versus Paul. Then any pains happening to TomPaul on that day hurt both persons.
But one more revision might do it: “Necessarily, only one person-stage can be hurt by a pain at any given moment.” With a careful definition of person-stage, emphasizing (as you say) that they are discrete systems, that can probably work. If this necessary exclusivity needs explaining, the job is already done, because you have made it true by definition, in defining “person-stage”.
I should have emphasized above that it might be one thing to identify objective colors and another to identify subjective color-sensations. I think it is – I have a thought experiment about temperature sensations that’s relevant here.
“although any experience must always be given from a first-person perspective (otherwise it is not an experience), it can also be given from a second-person perspective in empathy. To deny this claim commits one to the view that we never experience the thoughts and feelings of another person in any sense, but can only infer their existence from physical behaviour. This view involves a highly problematic dichotomy between inner experience and outer behaviour.” Evan Thompson [2004]. I am quite partial to much of what the enactivists say. I don’t see so much of a difference between the discussions of laws of physics (like is the trajectory of a billiard ball a real thing?) and laws of behaviour (is my trajectory from bed out to the loo a real thing).
Paul, since pains are picked out by their being painful, I don’t think what the neurologist sees on the screen – what she is visually experiencing as she looks at the image – is my pain but rather an image of its neural correlates. In general I’d say we’re never in an observational relation to experiences (not even our own experiences) since they aren’t possible objects, but rather the terms in which objects appear to us in our being conscious.
I take your point about TomPaul, but that doesn’t challenge the subjective privacy of experience for the fused person who later divides; it doesn’t place experience in the public, intersubjective realm of brains and bodies, what I call “represented reality.” Speaking of closest continuers, if you’ve not seen my paper on death which uses a Parfitian thought experiment to defeat the intuition of impending nothingness at death, you might enjoy it, https://www.naturalism.org/philosophy/death/death-nothingness-and-subjectivity .
Thanks for your thought experiment, I didn’t know you published at Less Wrong, cool!
The experience, of one’s own thought verses the experience, of-about one’s own thought…Irresistible!
My explanation of, experience, from today’s visit to my dentist…after my lower left rear jaw and left ear were numbed…
I experienced an itch there, I touched the area but there was no sense of touch, I wondered, what is consciousness…is an itch representative of instinctive consciousness, while touch is representative of functional consciousness, while my partially being there became an object of observation…
Then, is consciousness always in place in life, available for observation…
…can life, my life, become more objectifiable…
David (#17), seems to me that my empathy for someone else’s experience, for instance the pain I believe they are undergoing, is strictly my experience, not theirs, and their pain is strictly theirs, not mine. My belief and the resulting empathy comes from seeing their behavior, not from experiencing their experience, which seems to me an impossibility. Behavior correlates reliably with experience, so I don’t see any problematic dichotomy, rather it’s how I can pretty reliably tell what sort of experience someone is undergoing.
Tom,
I think of “see” and other perception verbs as crucially representational. To see something or feel it or smell it is to register that item via that sense. I am quite happy to say that a well trained physicist sees an electron in a cloud chamber (provided there was an electron), and not just a string of bubbles or droplets. In fact her indifferent response to different methods (bubbles vs droplets) of visualizing particles is further evidence of what is seen. Despite my fondness for a reasonably lightweight notion of qualia, I’m not willing to put cognition behind a veil of sense data. But this leads off into epistemology, probably beyond the scope of this discussion.
Tom, at #20…one of your examples is “the sweetness of sugar as I experience it”, which presumably you agree is the experience that the pastry chef tweaks for herself, but also expects will be shared by people eating the dish she has made. Art in general is aimed at manipulating the qualia of others.
The next bit of the Thompson quote is “In face-to-face encounters, we are confronted neither with a mere physical body, not with a hidden psyche, but with a unified intentional whole”. I don’t if that really helps me understand how this happens, but I suppose good physicalists think that this will be eventually uncovered by the neurophenomenalist program.
Paul, agree about not putting cognition behind a veil of sense data. We see objects, not experiences of objects. Still, experiences are the terms in which we see them – reality always appears to us situated, limited knowers in terms of some sort, whether conceptual, propositional, quantitative, or qualitative. I call this pretty uncontroversial claim “epistemic perspectivalism.”
David, yes the chef rightly supposes each customer will experience her pastry as tasting sweet but my point about the privacy constraint on consciousness is simply that each individual customer’s experience isn’t shareable. It’s one to a customer. Physicalists want to somehow put or find experience in the public domain, but I don’t think that’s in the cards.
If you think about the experiment where people have chop sticks gently poked on their back and at a certain point they can’t tell if it’s one or two, that gives a model for understanding consciousness that is just derived from the physical. Okay, if the human back is like that, what about the human brain? What if at a certain point the brain is unable to distinguish things as being two and instead treats them as one?
In terms of representation, when the number of chop sticks can’t be determined, you basically have a virtual reality sense of what is going on. Something is poking the back, but the exact detail of it is missing – thus you’re dealing with a virtual reality.
So consciousness is a virtual reality, created from where sensory input (including how the brain senses other parts of itself) isn’t high enough resolution to fully detect what is actually there. Thus you get a fantasy world with a high number of correlations with the real world. Enough to make it not seem like fantasy – and given the inability of the senses to detect certain resolutions of reality, no one has any better insight into the nature of reality so it is treated as reality. Until science comes along.
peter
I just don’t see how or where the idea comes fron that conscious experience has anything to do with representation. This is more indirection fallacy.
What conscious experience gets generated, gets generated. There is no way we can know if it ‘looks like’ anything – or indeed if there is any meaningful way we can describe what conscious experience is as ‘appearance’, when there is no objective sense of “appearance” outside of the scope of conscious experience in the first place.
There is also a huge amount of implicit assumption taking place in these theories. Just because something is outside of the scope of physics doesn’t mean it’s not real. It doesn’t even mean it’s not “physical”, in the generic sense. It just means that physics can’t deal with it – as could be expected for any human art.
Experience is as real as a spaceship and is no less real for having no account on physics. If having an account in physics is the hallmark for “reality” then we really are in trouble, as physics is a highly limited syntax-only discipline with an extremely narrow scope. As far as mental experiences go, it’s utterly useless. Then again – same with biology, where physics is also totally useless.
JBD
Tom,
I agree with epistemic perspectivalism, but don’t see how it helps your case. You’d need something more to block the ability of the well trained neurologist to see your pain on the fMRI. I mean, sure, she has her own visual qualia, which are distinct from the distal object of her perception, your pain qualia. But in every sense-perception, there are proximal qualitative properties that are separate from, but still don’t block access to, the distal events or processes. Nor is it generally required that the observer’s mental state contain the defining quality (here, painfulness; in an objective case, say squareness) of what is observed.
Paul: “You’d need something more to block the ability of the well trained neurologist to see your pain on the fMRI”
The well-trained neurologist could perhaps reliably tell when I’m in pain, but wouldn’t literally see my pain, which is picked out by the sensation, not an fMRI, which we’re assuming is reliably correlated with my reports of the sensation. My pain would have to be public for it to be the distal object (or event or process) of her perception – for her to have access to it – but what she has access to, what’s public, is my brain and the fMRI, not my qualia. If my pain qualia were intersubjectively accessible, we wouldn’t need the neurologist to interpret the brain scan for us non-experts as evidence of my being in pain.
Tom,
If I had a good theory of perception and its intentionally, I’d deploy it to see which of us was right about what neurologists could see. But I don’t.
john davey, totally in agreement with your post. Do you have a website or a forum and if so would you link it here? Thanks.
Hi Jeanne, just fyi the paper referred to in this post was ultimately published in the Journal of Consciousness Studies, the final draft is at https://naturalism.org/sites/naturalism.org/files/Locating%20Consciousness_0.pdf
Hello Tom. Thank you for the link. I’ll read the paper today.