In previous posts I gave arguments against the apparently widely held belief, at least amongst readers of this blog, that all mental states and phenomena, consciousness in particular, are reducible to computational processes in the brain: computationalism. Whether this is or isn’t the case divides the philosophical community, with famous names agreeing (Searle, Penrose) or disagreeing (Dennett, Chalmers).
This post is not another argument against computationalism but an attempt to be more scholarly. I will take a brief critical look at one of the more influential papers in the area, Chalmers’s “A Computational Foundation for the Study of Cognition“ (Journal of Cognitive Science 12:323-57, 2011). There are some things I admire about this paper. For once, it sets out quite meticulously to define what a computation is, making the important distinction between abstract computation and implemented computation. Even though as a computer scientist I find many points of disagreement in the detail of how this theory is constructed, the fact that Chalmers considers it in the first place is admirable. Many computationalist philosophers fail to make this distinction, which renders some of their arguments sloppy and implausible. But at least Chalmers answers definitively some rather silly (at least to a computer scientist) questions raised by Searle and Putnam about the relationship between computation in the abstract and the concrete. The opening sections of the paper found me nodding in agreement most of the time.
The main thrust of Chalmers’s computationalist explanation is in Section 3 which spells out his theory of “organizational invariance of mental properties”. Unlike the previous section, where technical words are used in their technical sense, here technical-sounding words are used in a rather vague and informal way. Sentences like this set off all sorts of alarm bells, as the amount of sense they make is questionable: “Causal topology can be thought of as a dynamic topology analogous to the static topology of a graph or a network.” I shall not quibble with language though but try to assign sensible and charitable meaning to the argument in order to make a sensible counterpoint.
The argument is conveniently summarised by Chalmers, so I can quote it in its entirety:
The argument for this, very briefly, is a reductio. Assume conscious experience is not organizationally invariant. Then there exist systems with the same causal topology but different conscious experiences. Let us say this is because the systems are made of different materials, such as neurons and silicon; a similar argument can be given for other sorts of differences. As the two systems have the same causal topology, we can (in principle) transform the first system into the second by making only gradual changes, such as by replacing neurons one at a time with I/O equivalent silicon chips, where the overall pattern of interaction remains the same throughout. Along the spectrum of intermediate systems, there must be two systems between which we replace less than ten percent of the system, but whose conscious experiences differ. Consider these two systems, N and S, which are identical except in that some circuit in one is neural and in the other is silicon.
The key step in the thought-experiment is to take the relevant neural circuit in N, and to install alongside it a causally isomorphic silicon back-up circuit, with a switch between the two circuits. What happens when we flip the switch? By hypothesis, the system’s conscious experiences will change: say, for purposes of illustration, from a bright red experience to a bright blue experience (or to a faded red experience, or whatever). This follows from the fact that the system after the change is a version of S, whereas before the change it is just N.
Before pointing out the logical fallacy of this argument (can you spot it?) lets travel in time, to a strange and dystopian future where in the wake of an untold cataclysm humanity will have lost some of its scientific knowledge. Among the forever lost items is all knowledge of radio communication. So when the archeologists of the future discovered a radio set in working condition it caused somewhat of a sensation. They surrendered the radio for examination to their contemporary scientists and philosophers. Moreover, unbeknownst to all, one fully automated nuclear-powered radio station, hidden among the ruins of the past, was still eternally broadcasting.
The scientists of the future figured out quickly that the radio was an electronic device, plugged it in and delicately experimented with the various knobs and buttons. The on/off button and the volume knob where easy to figure out. Their initial theory was that they discovered a white-noise machine. The philosophers of the day explained at length and convincingly why ancient folk might have wanted to build such a machine. But their speculations where quickly abandoned when the scientists figured out that the most mysterious knob, which we call the dial, had a function after all: If you point it to a particular number on the scale the device starts producing music instead of noise. Being able to listen to what was probably the sacred music of the ancient culture caused widespread fascination. Historians established with some confidence that Madonna was none other than the mythical figure revered as the mother of God in the ancient religion of Christianity.
This was a most interesting discovery not only culturally but also technologically. In the dystopian future digital devices such as computers and digital music players were still manufactured. So there was a solid understanding of music-producing electronic devices. The scientists figured out rather quickly that the radio was not digital, but analog. However, its input-output behaviour was precisely that of a digital music player, so they assumed it was some kind of analog equivalent of a digital music player. They even figured out what the dial was: a broken shuffle knob.
But they were not quite sure, and that was really annoying them. They knew that the ancient civilisation had mastered some lost technologies so they kept wondering. Was that really a music player, just analog? One of their prominent philosophers argued persuasively that ‘yes’:
All the components of the radio are electronic, but analog. So you could isolate each electronic component, interface it to the rest of the device via a kind of modem, then reimplement its functionality digitally. When two digital components are connected we can just remove the superfluous modems. So, in a step-by-step way, we can replace the entire analog device with a digital device so that we end up with a digital music player.
Persuaded, most scientists gave up further investigations into the nature of the radio. Except one well funded and talented scientist who questioned the argument and set out to actually put it in practice. Painstakingly, she figured out and replaced the analogue components of the radio with digital components — until she reached the antenna of the radio. She was shocked by the behaviour of this component. It was a very basic-looking arrangement of conductive elements which produced output out of apparently nothing. The only plausible hypothesis was that this was the memory element of the device, which stored all the music coming out of it. But it was much too simple in structure to be a memory. Our scientist was baffled. She wrote a few research papers about the stupendous analog memory technology of the ancients, but they got a mixed reception. The tabloid magazines made quite a stir around this mysterious technology but the established research publications refused to publish such speculation.
We are not sure what happened next. Did our scientist give up? Did she eventually discover electromagnetic waves and unlocked the true mystery of the radio? Most likely her lack of publications resulted in her losing her research funding.
Lets return to Chalmers now. His recipe for building a silicone computer out of a brain is similar to the above recipe for building a digital music player out of a radio set. The fallacy of the argument is called begging the question: when you assume that all brain components can be replaced with little silicone computers you already assume that all relevant functionality of that brain component is of a computational nature. But this is what the argument set out to prove in the first place, that the activity of the brain is of a computational nature.
If not computation, then what else is there in the brain that could realize mental states, in particular consciousness? Perhaps not radio-waves either. My money is on the theory that we don’t have the faintest clue. Just like our talented scientist from the future could not imagine radio waves, so we cannot understand what is going on, really, in the brain. Just like her, we understand a lot about various phenomena in the brain — except for the essence of how and why it works at a fundamental level. Our attempts to explain it using whatever concepts we have at hand will be, I have little doubt, as laughable as the miasmatic theory of disease or the phlogistonic theory of combustions. I admit, my theory is not a fascinating one.
Other than that, the paper is nice and readable. With this (fallacious) argument in place and its (mistaken) conclusion reached everything else fits neatly together. Just don’t use it as a recipe if you are trying to build a digital radio.