The sincere art of obfuscation: A rebuttal of Keith Frankish

Vassily Kandinsky's 'segment blue,' 1921 (cropped).
Could it be that your experience of seeing these rich colors doesn't actually exist?

After my publication of a rebuttal of Michael Graziano's latest essay yesterday, a Twitter exchange followed with philosopher Keith Frankish. It turns out that Frankish holds a very similar position to Graziano's: he, too, argues that subjective experience, phenomenality, is an illusion. This is called the 'illusionist' position in philosophy of mind. In the online exchange, Keith invited me to point out what is wrong with his ideas, as expressed in an online essay on aeon Magazine:

Not in that sense, I believe. I try to explain in the article. Where do you think I go wrong?

— Keith Frankish (@keithfrankish) September 26, 2019

As most of you know, I have little respect for the illusionist view, considering it self-evidently absurd. Well, in all honesty, I actually don't have any respect at all for it. But it is legitimate for Frankish to ask me to point out, explicitly, where I think his argument goes wrong. After all, having criticized his position publicly, I feel obliged now to be quite specific and explicit about my points. Moreover, Frankish writes in a sober tone and articulates his argument fairly carefully. The attitude he brings to the debate renders his work deserving of careful consideration, at least one time. So here we go.

Conflation

While setting up the context of his argument, Frankish correctly highlights a premise of the physicalist metaphysics, but incorrectly conflates it with science:

For science tells us that objects don’t have such qualitative properties, just complex physical ones of the sort described by physics and chemistry. The atoms that make up the skin of the apple aren’t red.

I don't think science says this at all. Instead, it studies and describes the behavior of nature. As such, it doesn't make—and fundamentally cannot make, as its empirical methodology cannot address such questions—assertions about the metaphysical status of any properties. Science simply describes the behavior of objects and phenomena as they appear to our observation. Such descriptions entail measurable physical and chemical quantities, but that doesn't entail or imply a metaphysical exclusion of qualities from nature.

Having said that, it is entirely true that the metaphysics of physicalism is premised on the notion that all qualities are generated by the brain and, as such, cannot exist out there in objects, but only inside our heads. Frankish's assertion quoted above is consistent with his physicalism, but it illegitimately co-opts the success of science as if physicalism were implied by it. While a common move, this is wrong.

Question-begging

Insisting on his conflation of physicalism with science, Frankish claims:

It is phenomenal consciousness that I believe is illusory. For science finds nothing qualitative in our brains, any more than in the world outside. The atoms in your brain aren’t coloured and they don’t compose a colourful inner image.

He elaborates beyond the quote above, but the complete essence of his point is already captured in it. The argument structure is this:

1. Physical things, in themselves, have no qualitative properties (like color, flavor, tone, etc.). Only our perceptions of them do;
2. The brain is a physical thing;
3. From (1) and (2), the brain has no qualitative properties;
4. Our experiences are reducible to our brain;
5. From (3) and (4), our experiences cannot entail qualitative properties.

Ergo, qualitative properties—phenomenality, subjective experiences—cannot exist; they must, instead, be an illusion. The question-begging here is rather obvious: step (4) in the argument structure above presupposes the metaphysics of physicalism, which is precisely the point in contention.

Ironically, what Frankish actually accomplishes in his argument structure is to highlight an implication of physicalism that reduces it to absurdity.

More question-begging

Frankish proceeds to argue against two alternative metaphysics: property dualism—the view that the brain has both physical and qualitative properties—and the view that qualitative properties are merely how physical properties present themselves to introspection, being, therefore, ultimately just physical.

He rejects property dualism by arguing that the physical world is causally-closed and, therefore, the additional qualitative properties are useless and can presumably be dismissed on parsimony grounds. Then he rejects that qualitative properties are merely appearances of physical properties because

it is not just that introspection fails to present sensations as brain states; it positively presents them as utterly unlike brain states

I concur with Frankish's conclusion that both alternatives are incorrect, even though I'd be a lot more cautious than him about claiming that the physical world is causally-closed. At a microscopic level, all quantum mechanical events are undetermined. Only at a macroscopic, statistical level do regularities emerge that allow us to speak of causality. Moreover, laboratory experiments, by virtue of their very need to isolate experimental conditions from unknown factors, may exclude non-local organizing principles in nature that may not be describable as physical causality (cf. e.g. this).

Be that as it may, my point here is different: all alternatives considered by Frankish assume physical realism; that is, the notion that there are non-experiential things out there. This is a premise of physicalism and certain variants of panpsychism, but not of other metaphysics. Objective idealism, for instance, while granting that there is indeed a world out there, maintains that such a world is itself constituted by transpersonal phenomenal states. These transpersonal states simply present themselves to us as the qualities on the screen of perception, in a qualitative transition that occurs for reasons I've discussed on Scientific American. This completely avoids the impossible transition from one ontological category to another, as the fact that certain qualities of experience modulate other qualities of experience is empirically trivial (it happens e.g. every time your thoughts affect your emotions, or the other way around). Finally, a very strong case can be made that physical realism has already been refuted by experimental physics anyway, as I've discussed also on Scientific American here and here.

Whatever the case, Frankish's argument begs the question of metaphysics by simply assuming a key premise of physicalism contested by other metaphysics. At best, his argument refutes other variations of physicalism, but says nothing about e.g. objective idealism.

Internal inconsistency

By rejecting that qualitative properties are introspective appearances of the physical brain and taking physical realism as a given, Frankish concludes that only illusionism can be true: introspection misrepresents the physical states of the brain, thereby generating the illusion of qualitative properties. We've already seen above how the path he took to arrive at this conclusion begs the question in more than one way. I shall argue now that, in addition to this, Frankish's elaboration is also internally inconsistent.

To begin with, I can't resist pointing something out that has already been pointed out by many others. Consider this passage by Frankish:

Think of watching a movie. What your eyes are actually witnessing is a series of still images rapidly succeeding each other. But your visual system represents these images as a single fluid moving image. The motion is an illusion. Similarly, illusionists argue, your introspective system misrepresents complex patterns of brain activity as simple phenomenal properties. The phenomenality is an illusion.

Frankish is very clear that what his argument tries to deny is the very existence of qualities, experience, phenomenal states. But since illusions are themselves phenomenal states—after all, they are experienced—they are already instances of the very thing whose existence Frankish is trying to deny. The appeal to illusions immediately disproves Frankish's whole point. He explicitly addresses this objection towards the end of his essay, and I will deal with his answer towards the end of mine, in the last section below.

For now, though, let us charitably interpret the reference to illusions as a metaphorical effort to evoke a certain familiar intuition, and see where Frankish goes with it:

it is useful to us to have an overview or ‘edited digest’ (Dennett’s phrase) of [our brain] processes – a sense of the overall shape of our complex, dynamic interaction with the world. When we speak of what our experiences are like, we are referring to this sense, this edited digest.

The point here is that, when we introspect, what we experience aren't the original brain processes as they are in themselves, but an inaccurate, distorted, "edited digest" of these processes. This is the basis for Frankish's claim that experiences are illusions: they are misportrayals of that which they represent, i.e. physical brain states.

There is at least one obvious problem with this, though. Misportrayals as they may be, since Frankish's basic premise is that only physical states exist, these 'edited digests' must themselves consist of physical states; what else could they be? And so we end up with the exact same question we started with: How is it that these latter physical states—i.e. the brain states corresponding to the misportrayals—are presented as something utterly unlike brain states?

It seems to me that, to answer this question within Frankish's own logic, we need to postulate a meta-introspective system that misportrays the misportrayals. But then such meta-misportrayals will also necessarily consist of physical brain states, for there is nothing else they can consist of under Frankish's premises. So we need a meta-meta-introspective system that misportrays the misportrayals of the misportrayals... Well, you get the picture.

Frankish's effort to add a physical layer of indirection to explain how presumably physical states can present themselves as qualitative properties is like trying to pull yourself up by your own bootstraps; it just can't do the magic Frankish wants it to do. Adding the indirection brings things no closer to a solution; it merely ends up confronting the exact same problem—intact—that it started with.

No amount of physical indirection can make the physical seem phenomenal, just as no amount of extra speakers can make a stereo seem like television; these two domains are incommensurable. All Frankish accomplishes with his step of indirection is to postpone the inevitable, final confrontation with the real problem at hand. Yet, by obfuscating the innate simplicity of the issue, these indirections can create the impression that some profound, penetrating philosophical insight lies hidden behind them. But none does; it's all smoke and mirrors, as I shall argue in more detail below.

More internal inconsistency

Even Frankish's chosen metaphor actually illustrates no more than the untenability of his thesis:

In Consciousness Explained (1991), Dennett draws a comparison with a computer’s user interface, with its icons for files, folders, waste basket and so on. This is a fiction created for the benefit of the user (a ‘user illusion’). By manipulating the icons, we can easily control the computer without knowing anything about its programming or hardware. Similarly, representations of phenomenal properties are simplified, schematic representations of the underlying reality, which we can use for the purposes of self-control. We should no more expect to find phenomenal properties in our brains than to find folders and waste baskets inside our laptops.

Let us interpret this strictly according to Frankish's own premises and logic, so as not to misrepresent his case: the user interface (UI) is a fiction that (mis)represents e.g. computer files. The latter are patterns of open and closed microelectronic switches in a silicon memory chip inside the computer. But they are presented to the user, through the UI, in the convenient form of little icons. The files aren't icons—they are patterns of open and closed microelectronic switches—yet the UI's (mis)representation is convenient for the user.

So far so good. One can go further down this line of reasoning and observe that, just as the actual computer files, the UI, too, is purely physical: the pattern of pixels that makes up the icons on the screen also consists of open and closed microelectronic switches in both a memory chip inside the computer and the LCD screen that displays the icons to the user. In other words, the UI example shows that a first set of physical states (the actual files) is (mis)represented by a second set of physical states. The states in the second set are different from the states in the first set, which accounts for the fact that icons are different from the actual computer files; but all states are physical and will never look like anything other than physicality.

Transposing this to our problem, a first set of brain states—corresponding to e.g. brain signals processing visual information—is (mis)represented by a second set of brain states, the latter playing the role of UI. These sets differ in that the brain states comprised in them differ. But they are all still brain states; they are all still physical. There is nothing in Frankish's metaphor that provides any intuition for how something physical can end up looking like something phenomenal.

Indeed, the metaphor only seems cogent because it cheats: its evocative power rests in the transition from abstract physical states hidden inside a computer chip to the experience of seeing the computer screen with its icons. But by visualizing this transition we are already using that which Frankish claims not to exist: phenomenal states, qualitative properties, experiences. To be strictly consistent with Frankish's logic, we must imagine that no one is there to look at the computer screen. Then, we are left only with the physical states inside the computer chip and those of the LCD screen. There are no qualitative properties anywhere, only physical states. Now, without someone to look at the screen, does the metaphor do what Frankish wants it to do?

You see, the unintended cheat—for I believe Frankish is cheating himself too, insofar as he sincerely believes his own argument—is that the metaphor implicitly appeals precisely to the very thing whose existence Frankish wants to deny. Therein resides its entire evocative power. Once you see it, the metaphor not only collapses, but its meaning also reverses: no amount of physical representation of the physical can create the appearance of phenomenality.

Some more commentary

Frankish begins now to conclude his argument:

If we observe something science can’t explain, then the simplest hypothesis is that it’s an illusion, especially if it can be observed only from one particular angle. This is exactly the case with phenomenal consciousness.

Except that, in the case of phenomenal consciousness, an illusion is already an instance of phenomenal consciousness, the very thing Frankish denies. Moreover, there are other metaphysics that place the observable dynamisms, patterns and regularities of phenomenal consciousness firmly within the framework of science (see e.g. my own work here, which has been summarized in a popular essay on Scientific American). Therefore, the claim that we have to deny phenomenality because "science can't explain" it is completely bogus; it arises merely from an apparent inability to look at the problem from a different angle, with at least fewer unexamined assumptions.

Frankish seems to be so closed up in his box of implicit assumptions he can't see any alternative but to deny the most obvious. Consider this long paragraph, which Frankish presents as a reason to believe in illusionism. I will quote it in full because I find it so remarkable:

A second argument concerns our awareness of phenomenal properties. We are aware of features of the natural world only if we have a sensory system that can detect them and generate representations of them for use by other mental systems. This applies equally to features of our own minds (which are parts of the natural world), and it would apply to phenomenal properties too, if they were real. We would need an introspective system that could detect them and produce representations of them. Without that, we would have no more awareness of our brains’ phenomenal properties than we do of their magnetic properties. In short, if we were aware of phenomenal properties, it would be by virtue of having mental representations of them. But then it would make no difference whether these representations were accurate. Illusory representations would have the same effects as veridical ones. If introspection misrepresents us as having phenomenal properties then, subjectively, that’s as good as actually having them. Since science indicates that our brains don’t have phenomenal properties, the obvious inference is that our introspective representations of them are illusory.

For all I know our phenomenal properties—i.e. our subjective experiences—indeed do misrepresent something, either physical states or other phenomenal states corresponding to the world outside or certain aspects of our body. But even then they are still phenomenal. One can't deny phenomenality merely by arguing that phenomenality misrepresents something, for this presupposes the phenomenality that misrepresents something.

Rebutting rebuttals

Frankish then begins to preemptively answer possible objections to his thesis. He starts with the objection that our knowledge of the world begins with consciousness, and so consciousness cannot be an illusion. He argues against this by saying that a simple robot would have only sensors and actuators, and only more sophisticated robots, evolved from the simple one, would develop a meta-cognitive introspective system like consciousness. He claims that the same applies to us, so consciousness is not primary but evolved.

One of many problems with this hand-waving argument is that phenomenal consciousness does not need introspection to exist; by assuming that phenomenal consciousness is restricted to its introspective mode, Frankish already makes a mistake. I elaborated on this in both a technical paper and, in summarized form, in a Scientific American essay.

The gist of the point is this: introspection—our ability to know and report that  we have an experience—is a metacognitive configuration on top of phenomenal consciousness proper. We know through e.g. the no-report paradigms of modern neuroscience that there can be phenomenal states beyond the field of metacognitive introspection. These states are experienced, even though subjects do not know that they experience them, and so cannot report them; not even to themselves. Once one sees that phenomenal consciousness is in fact more basic than introspection, Frankish's argument here, which is already hand-waving to begin with, collapses.

The grand finale

In answer to the objection that phenomenal states cannot be illusions insofar as illusions are themselves phenomenal states, Frankish has this to say, as a kind of grand closure of this argument:

This looks like a serious objection, but in fact it is easily dealt with. Properties of experiences themselves cannot be illusory in the sense described, but they can be illusory in a very similar one. When illusionists say that phenomenal properties are illusory, they mean that we have introspective representations like those that we would have if our experiences had phenomenal properties. And we can have such representations even if our experiences don’t have phenomenal properties. Of course, this assumes that the representations themselves don’t have phenomenal properties. But, as I noted, representations needn’t possess the properties they represent. Representations of redness needn’t be red, and representations of phenomenal properties needn’t be phenomenal.

I find this passage truly remarkable, but not for the reasons Frankish would presumably like me to. Let's dissect it: Frankish begins by acknowledging that "properties of experiences themselves cannot be illusory in the sense described." This seems quite final to me: the sense described suffices to prove that experiences themselves exist, even if "they can be illusory in a very similar" but other sense. If the sense in which experiences themselves must exist suffices to show that they do exist, whatever other sense in which they may be said to not exist is irrelevant to the point in contention. But let's proceed and see where Frankish takes us.

The sentences that follow are an unsurpassed accomplishment in presumably well-meaning, sincere, but tortuous obfuscation and confused thinking. You should not feel bad if you can't make heads or tails of them, for I had to re-read them several times to see where Frankish is trying to go. What he is saying is that, whether we have actual experiences—phenomenal properties—or not, everything happens as if we had them. That he thinks this answers the objection baffles me, for it in fact succumbs to the exact same objection: for things to happen as if we had experiences, it must seem to us as though we did have them, even if we don't. But Good Lord, the seeming is already an experience. The introspective representations must themselves be phenomenal, otherwise there would be no seeming. Yet there obviously is seeming, for what is an illusion but a factually wrong seeming? If he thinks there is no seeming, why is Frankish trying so hard to convince you that what seems to be the case actually isn't?

Frankish is tying himself up in knots to somehow avoid what is obvious to just about everyone else. It is remarkable and at the same time painful to follow his argument as he buries himself in conceptual confusion. That he claims that the original objection has been "easily dealt with" in this manner is ironic to say the least.

And then he admits:

But how does a brain state represent a phenomenal property? This is a tough question.

Oh! All right!

You see, under Frankish's premises, the question isn't "tough;" it is by construction impossible: for him there are no real phenomenal properties; it just seems as though there were. This seeming is created by said brain representation or state, which misportrays other brain states. Now, how can a brain state create the seeming if seeming—i.e. phenomenality—is not allowed to begin with?  Talk about internal contradictions and conceptual confusion...

Frankish's entire case rests on at least a tentative answer to the question above; without it, there is nothing, just smoke and mirrors. But he just says it is "a tough question"... Oh well.

Undeterred, as if he had accomplished anything at all with everything he has said thus far, Frankish continues:

I think the answer should focus on the state’s effects. A brain state represents a certain property if it causes thoughts and reactions that would be appropriate if the property were present.

Blatant question-begging again. Only under physicalist premises could effects sufficiently account for the question Frankish is leaving open. What defines phenomenal states is precisely that, regardless of their effects, there is something it is like to be in them. By claiming the above Frankish is arguing circularly. But he goes on:

I won’t try to develop this answer here.

Only the answer to this question is substantive for the argument he is trying to make. By not even trying to answer it, Frankish rests his entire case on pure hand-waving. He argues that

it is not only illusionists who must address this problem. The notion of mental representation is a central one in modern cognitive science, and explaining how the brain represents things is a task on which all sides are engaged. ... There is a challenge here for illusionism but not an objection.

I find this nonsensical for a very simple reason: yes, everybody has to account for representation; but only illusionists have to account for it by acknowledging that we seem to have experience while denying experience. How can a physical state be 'seeming' when there is no seeming?

This is what makes their case impossible. Nobody else faces the same problem. An objective idealist, for instance, must account simply for how certain phenomenal states represent other phenomenal states. There is no ontological bridge to be crossed and thus nothing fundamentally difficult about it: our thoughts can trivially represent our emotions by e.g. naming and describing them. An eliminativist—i.e. one who denies experience without even bothering to account for some kind of 'illusion of experience'—only has to show how some physical states represent other, different physical states, which computers do all the time by the use of variables and pointers. The claim that everybody faces the same challenge here is simply untrue, and rather obviously so.

Frankish has accomplished precisely nothing in his long essay; at least nothing more than tortuous obfuscation and hand-waving.

Final thoughts

I started writing this essay with the sincere intention to be charitable, open and understanding. But I finish it now with the overwhelming impression that I have been commenting on a charade. Quite honestly, sincere as Frankish's effort may be, I do think the whole thing is indeed an outright charade, a farce, veiled in conceptual complexity and obfuscation.

Yet I don't think Frankish and other illusionists are malicious about it (perhaps it would do them less dishonor if I thought they were). I think they themselves are caught up in their own charade, drinking their own Kool-Aid to a degree they don't even suspect. Illusionism and eliminativism, in my view, are veritable psychological case-studies on how the human mind finds baffling ways to deceive itself so to defend its own prejudices. I say this with absolute sincerity, in that I truly believe it; it's not to sarcastically deride anyone. I just can't, for the life of me, fathom how otherwise intelligent and educated people can tie themselves up into so much sheer nonsense.

Trying to elucidate this psychological conundrum is, perhaps, the real discussion to be had, after all.