writing

> if not god, then what?
> fables for a young skeptic
> the thinker's toolbox
> the glass bead game
> fight dogma, not religion
> neural rhythmicity...

teaching

> beauty and the brain
> cyborg millennium
> einstein's universe

about

> resume / cv

Listener programmability, islands of truth, and the hard problem

Unpublished

Abstract

The judgment of whether a physicalist account could include all the facts of mental experience depends on whether the person receiving the account might be suitably influenced by it. In theory, a listener could possess an interface though which an account, e.g. of color vision, activated photoreceptors. In that case the account would produce in the listener all the necessary qualia: no facts would be missing.

Chalmers’ “hard problem” is whether a complete scientific account of neurophysiology, and its interaction with the world, would explain everything about the mind. Jackson’s widely cited diorama for this problem is of the scientist Mary, who, despite living since birth in an entirely black and white environment, possesses a functional color vision system. Would Mary know what it was like to see blue, providing she was provided a complete scientific account of color vision? Chalmers et al. say no. Dennett says yes, and I agree, but with the caveat that it depends entirely on how said account affects her.

Suppose that Mary's brain were arranged in such a way that the words, diagrams, and equations provided in the account were able to excite – by whatever process – the cone cells in her retina. Obviously, in that case, Mary would perceive color, since the excitation of those photoreceptors is sufficient for her perception of blue in more ordinary circumstances. In this case we would say that the account did give her the quale.

The normal human brain, however, is not configured in this way. There is no known way in which a linguistic or symbolic stimulus can excite photoreceptors. For Mary to really know what it is like to see color, something further down the line in the visual system would have to be activated by the account in just the same way as it would be activated by really seeing blue. Unfortunately, the same argument that applies to photoreceptors probably also applies to the lateral geniculate nucleus (stage 2 of the visual system), primary visual cortex (stage 3), and probably also many subsequent steps into both the dorsal and ventral visual processing streams. If our complete account of color vision is to impart to Mary the what-it-is-like phenomenology, it is going to have to do so at some late point in the perceptual hardware, and it is going to have to dance a potentially difficult number once it gets there.

Is there such a point? Can the mere words and symbols of a scientific account play puppeteer with neurons that adroitly? I don't know. I suspect not. It is not, however, impossible to imagine an affirmative answer to those questions. Suppose for instance that AI researchers succeed in building a software-based conscious being, called Ike, running on a computer at least somewhat resembling those extant today. Suppose that Ike passes the Turing test and even skeptics become convinced that Ike is not a zombie. Finally, stipulate that a mechanism is engineered into Ike by means of which any of his variables or algorithms can be modified via the provision of special spoken or written instructions. In that case, we could write the reductionist account of color vision, and/or engineer the interface to Ike’s internal variables (this is called an "application programming interface" or API in software lingo) to overcome what is, for Mary, a difficult barrier: the words, diagrams, and equations comprising the physicalist account actually would be capable of exciting Ike's "cone cells" or the variables with the same functional role. By virtue of that, and mutatis mutandis with respect to Mary, we could talk to Ike or give him a paper to read that would impart the actual experience of seeing blue. 

Ike demonstrates that in principle, a physicalist account of color vision, provided it is crafted for specific effect in a listener, would certainly lead to qualia. Whether this is also possible with Mary depends entirely on her receptivity to the message – in effect, her programmability. If Mary's mind/brain is such that she is insensitive to the account, then there's nothing for it – she just won't ever get the quale from the account. But this would also be true if she failed to understand – on a purely cognitive level – the science we provide her, or didn’t speak the language in which it was written. In such cases, the account’s failure to impart phenomenal experience would be Mary’s fault, not its own. There is no sense in talking about the adequacy of science to explain something unless we specify the effect that the communication of that science has, or could potentially have, on its recipients.

When framed this way it becomes clear that Chalmers’ hard problem is not new to the philosophy of mind. Maxwell's laws don't impart an understanding of classical electrodynamics, at any level at all, unless the person reading them is prepared to interpret them in the ways we care about. More broadly, this is a question about communication: if the sender of a message cannot reliably predict how it will affect the receiver, it will be impossible to send information about anything at all, phenomenal or otherwise. In some cases, such as Mary's but not Ike's, the receiver may entirely lack the API that would allow the sender to convey the desired information. It is the recipient's programmability that determines meaning, i.e. semantics, and it is the meaning that carries the force of an account. One may as well ask whether it is possible to turn a lump of coal into diamond just by speaking to it about differences in bonding structure. That this is not possible indicates not a weakness in the speech or incompleteness of its content, but the absence of an appropriate interface to the receiving lump's physics. Again – I don't know whether Mary has the required interface, but whether she does or doesn’t is a question about physiology, not philosophy or the in-principle completeness of physicalist theories of mental experience.

*     *     *

As with most things, there is some Wittgensteinian therapy to be done here, and Rudd (1997) has done so, albeit with a different purpose and outcome. His outcome was that the language of physicalism has a domain of application different from the language of phenomenology, and that the inapplicability of the former to the latter doesn’t mean that the objects of the latter don’t exist. They don’t exist as things, Rudd says, but this is not the same as saying that they are nothings, and he invokes Wittgenstein’s Private Language Argument to make the point. For hard-core physicalists like me (and perhaps Dennett too), this tastes a lot like Stephen Jay Gould’s description of science and religion as “non-overlapping magisteria.” Inadequacy by fiat seems like an appeal to ignorance fallacy, and in any case it gives language the upper hand over objective reality. I don’t think that’s what Wittgenstein would have wanted…but no matter, because he can help us here, in a way that Rudd just barely misses.

Any number of language games might be played in the description to ourselves of what it is like to be ourselves. We can work more or less hard on achieving internal consistency in such accounts, parsimony of expression, and even poetry that distills emotions and cognitions phenomenology out of the background noise. To the extent that others have developed internal languages that match our own, communication of such expressions may be received with considerable agreement, and when it does, our confidence grows that we have expressed something true.

The most important thing to remember about such expressions and their truth values, however, is that they operate entirely within a particular universe of discourse, i.e. that of phenomenology. (In this sense I agree with Rudd’s related point.) The objects to which expressions refer – any linguistic expressions at all – reside exclusively within our minds. There is an easy metaphor here to videogames. Modern videogames are so sophisticated that it is convenient to talk about one object being larger or heavier than another, because the software of the game simulates and recognizes things such as size and mass, and these attributes determine the interactions of objects within the game. Expressions about those attributes do have truth values, but they apply only within a universe of discourse, i.e. that of the fictional game world. There is no “real” world in which the objects of the game have a size or a mass at all; there are no objects in the first place. When we say the “real” world, we conventionally mean the world outside of ourselves, the world in which our bodies move, the world that contains the earth and the sun. In that world, the objects of the game simply do not exist as objects. They exist as something else, i.e. as patterns of bits inside the CPU of the game console or computer. But in that case – and this is the crucial point – our interpretation of the truth value of statements about the comparative masses of the game objects is quite different than the interpretation of similar statements made about real world objects.

It is much the same with statements about the phenomenological universe. We can sensibly and pragmatically discuss such things as whether an apple is red, but that discussion is about the phenomenological universe, not the “real” external one. We often ignore the distinction because it is convenient to do so, and because we evolved in settings where there was no reason to make it. But as Wittgenstein noticed, the rigors of philosophical discussion snap us out of our linguistic lassitude and make us attend more scrupulously to the real range of our statements.

If I say that I know what it is like to see the color red, and that that feeling differs from what it is like to eat strawberries, this is a perfectly sensible thing to say. Both statements are being made with respect to the same universe of discourse, i.e. the phenomenological one, and can be reasonably compared. But if I say that I know what it is like to see the color red, and that this feeling is not the same as some description of neurobiological dynamics, i.e. the neurobiology of color and object perception, then I am attempting a trans-universal comparison and am destined to achieve a nonsense answer or no answer at all. It is the same as if I asked whether a game object was more or less massive than a real world object, or if I complained that I knew – within the context of a game – what it felt like for an object to be too heavy to lift, and that this was not the same feeling as I had when I contemplated a description of the bits or bytes or subroutines that instantiated that simulated physics. Of course no explanatory parity could ever be achieved via such comparisons! The phenomena being described live in different universes. Neurobiology lives in the world of the earth and the sun, and my experience of redness lives in the world of idea. Plato was right, in a way. My reading of Rudd is that he comes within a hair’s breadth of saying this, but ultimately says something contradictory and anti-hard-core physicalist.

*     *     *

A certain breed of philosophers – Platonists, for instance – may insist, in spite of the above, on asserting the reality of the phenomenological universe. What is argued for above, they would say, is merely the incompatibility of explanations across universes of discourse – there is no dissolution of either universe. Such insistence, however, falls victim to the same trap but at a higher level: by asserting the “reality” of the phenomenological universe, one makes a claim about the existence of something. What do we mean when we say that something exists? What we usually mean is that that something is out there in the external world. We can always restrict the domain of our statements to some smaller space, e.g. “There is a rock troll in that cave” (referring to a game universe), but no such restriction is given here. The idea is that the objects of phenomenology exist, either in the same universe as the earth and the sun, or in some other universe that deserves equal status as real.

It is my central claim in this Wittgensteinian therapy session that no universe other than the one that contains the earth and the sun deserves membership in what we call reality. There “are” other universes, but they are only universes of discourse…they are isolated logical / language games with their own laws, their own constructs, and their own truth values. No truths that emerge from such islands constrain the universe in which such games are merely simulations – this is why Penrose’s arguments about Gödel provability’s implications for the physical basis of consciousness are unconvincing. Our inner experiential worlds are precisely such isolated universes...neverlands whose truths and facts obtain only as long as we stay inside them.

Wittgenstein could hardly have picked a more succinct or ambiguously powerful statement to open the Tractatus with than “The world is all that is the case.” Perhaps, if had he ever played XBox, he wouldn’t have recanted.

References

Dennett DC (2005) Sweet Dreams (Boston: MIT Press).

Dretske F (1995) Naturalizing the Mind (Boston: MIT Press).

Jackson, F. (1982), “Epiphenomenal qualia,” Philosophical Quarterly 32: 127–36; reprinted in Lycan (1990).

Nagel T (1974) “What is it like to be a bat?” Philosophical Review 83: 435–50.

Rudd AJ (1997) “What it’s like and what’s really wrong with physicalism: a Wittgensteinean perspective,” Journal of Consciousness Studies 5(4).

idea sketchbook

> colorblind
> probabilistic law
> general relativity
> hybrid cars