Yesterday, I had an excellent conversation with Peli Grietzer on the subject of qualia and the hard problem of consciousness, and the reasons why some attempts to defuse these problems (e.g., Dennett’s ‘Quining Qualia‘ or his response to Chalmers) leave people who are drawn to the seriousness of these problems cold. I recommend Peli’s work to everyone I know, and even some people I don’t. It’s simultaneously incredibly unique and incredibly timely, trying to unpick issues at the intersection of aesthetics, epistemology, literary theory, and philosophy of mind using the theoretical insights implicit in deep learning systems, particularly auto-encoders.
Peli has come the closest of anyone I know to giving real, mathematical substance to the most ephemeral aspects of human experience: moods, vibes, styles; the implicit gestalts through which we filter the rest of our more explicit and compositional cognitions. Why does being in Berlin have a distinct cognitive texture that’s different from being in London? It’s not any one thing you can put your finger on, but a seamless tapestry of tiny differences that you drink in without needing to reflect on it. So when Peli says that he’s not sure that these most immediate, holistic, and yet nonetheless cognitive features of our experience are being taken seriously, I have no choice but to listen, and to see if there’s another way to articulate my own worries.
This result was probably the best thought experiment I’ve ever managed to articulate, and probably the most concise explanation of my problems with certain styles of thinking in the philosophy of mind. Given that my thoughts here are also in some sense an engagement with Wittgenstein and his legacy, I thought it might be nice to try articulating them in an aphoristic style. Here is the result.
Here’s another logical thought, related to @lastpositivist’s idea of informal omega consistency (IOC). I think it can be used to explain the logic of mysterianism, specifically as applied to the ‘hard’ problem of consciousness and/or qualia.
So, IOC is about the dialogical structure of quantifiers: someone accepts that their is an instance of some type, perhaps with some particular property, but not only do they refuse to produce that instance, they reject every candidate presented to them.
The refusal is, logically speaking, non-constructiveness (it violates the existence property). But it is the rejection (call it obstructiveness) that is more sinister, and which demands the title omega-inconsistency, insofar as it is only inconsistent in the limit.
If we mix our registers, and supplement the dialogical approach with the substitutional approach, we can see how one can in fact name this mysterious instance (‘x’), while refusing to allow any other singular term to be substituted for it. ‘x’ only refers in the limit.
This explains my peculiar frustration with talk of ‘consciousness’ and ‘qualia’, which we might, respectively, call holistic and atomistic mental states/objects/properties/whatever. These terms are slippery, because their reference is almost always secured by acquaintance.
There are exceptions to this: e.g., the global workspace theory of consciousness (good), or even Sellars’s processual theory of qualia (bad). These theories furnish the terms with referents. However, there are many who act as if they want this, while rejecting it in the limit.
Those in the sway of the language of ‘what it is like to…’ are drawn to deixis is like a moth to a flame: ‘it is like this’, ‘you’re experiencing it now’, ‘it’s what is happening here’; as if they could point internally, at the hidden mystery of themselves.
Here’s the thing though: what if we could point internally, would even that satisfy them?
Let’s assume that something like the predictive processing/computational neuroscience view of the visual system (dorsal/ventral; V1-V6) is correct. Now say that I can use transcranial magnetic stimulation to modify the activation patterns in one or more of the feature layers.
For those not completely aware of the terminology here, the thesis is that there are layers of features going from concrete to abstract, each extracted from those below (e.g., edges > figures > objects > types, though more messy than this).
We can take the usual example of colour constancy (e.g., ‘this object is red [all over and persistently]’). It’s pretty obvious to us that there is information beneath this classificatory layer, that there are regions of the objects whose shades vary spatially and contextually.
Check out some colour constancy illusions if you want to attend to these lower level features, and see how it is that your visual system is extracting a sort of classificatory gestalt:
Now, say that I start randomly stimulating these feature layers. This obvious lower level information is going to be chaotic, but this chaos will be accessible to you, even if it doesn’t lead to meaningful higher level features such as colour constancy.
This is non-conceptual content. It is information. It has structure. And it is in some sense cognitively accessible by higher level processes. Indeed, if I asked you to tell me when your visual field strobed ‘red’, you could probably do so.
What if I then got you to use this information you have access to in order to guide my explorations through the space of possible sensory content? What if I asked you to tell me ‘closer’ or ‘further away’ when I was searching for a particular colour sensation?
Here I am asking you to use your cognitive access to lower level features of your sensory processing mechanisms to point me in the right direction. I imagine that some people will be much better than others at this.
An artist might be able to locate ‘chartreuse’ in their imagination more quickly and deftly than a philosopher. A chef might sniff out ‘cumin’ where a neuroscientist can only get as far as ‘spice’. This is a genuine form of knowledge, but it is knowledge in the sense of know-how.
It is this know-how that Mary lacks. She could have a dream in which she encountered something red in an abstract sense, or in which she encountered something red in the concrete sense, but she could never connect the two, nor lead you to that part of her internal feature space.
This is what is deceptive in the language of ‘knowing what it is like to…’. It tempts us to think that there is some special sort of knowing-that which acquaints us with a what, mediated by a use of the word ‘like’ divorced from all abilities to compare the concrete.
This is pretty much the only case in which I still think that something like the later Wittgenstein’s philosophical therapy is required, where the words we choose lead us astray. Though I reject the McDowellian form this takes, which always aims to explain away, rather than explain.
A better choice of philosopher here is Sellars, whose analysis of ‘looks’ talk in EPM remains prescient. All that it lacks is an analysis of how the wider vocabulary of ‘seems’ gets used as proxy for calibrating our capacities to monitor and modulate our sensoria.
Wittgenstein’s most serious error is his talk of ‘the beetle in the box’, because he radically underestimates the sheer extent of our ability to talk to one another about the contents of our minds, and guide one another to different parts of the space of possible cognition.
The history of art and literature should be enough to demonstrate this. What else does an expressionist take themselves to be doing when they communicate to us features of the human condition we have not yet felt?
So, when I put the mysterian in the chair, and let him point, when I exclaim ‘I have found it! I have found what it is like to see red!’ will he turn around and tell me, once more, that he cannot accept my substitution?
Here is how I think that mysterians will respond to this argument: they will cast any and every aspersion upon the veracity of the actual neuroscience that I have used to set up my thought experiment, appealing to some future science that in the limit will always know better.
But what if they are asked directly: could there be some real such future neuroscience, enabling a similar experiment, under which you would be happy to substitute your ‘what it is like’ terms for neuroscientific terms? I suspect they will react like conmen whose jig is up.
As for the McDowellians, they will flat out deny that any neuroscience, real or ideal, will ever have answers for us. They will insist that non-conceptual content is a contradiction in terms, and then, when we ask what conceptual content is, tell us to stop asking questions.
Mysterians and quietists united at last, in their mutual desire to ignore anyone actually trying to answer the questions they purport to be interested in.
There are various ways I could expand on this, but I’m quite happy with the concision. It’s a good example of the sort of philosophy I think can be done online that simply cannot be done in a journal article. Precisely the sort of stuff that Wittgenstein did, at his best, even if he was always irresistably drawn to the mystique of silence.