Wednesday, August 16, 2006

The Machinery of Explanation

Children start life unable to speak but quickly learn the language of their culture. It has been remarked that the actual contact time children have with a language does not in itself provide enough information for them to make sense of it from scratch. It seems therefore that newborn humans have an innate understanding of some basic linguistic constructs. These constructs act as template categories and the particulars of a given language rapidly fall into the template slots as children learn. In short, human beings have prior expectations about how language works and these expectations, if fulfilled, speed along the process of language learning. (See Pinker)

It is likely that humans also have prior expectations in a more general sense about the “syntax and semantics” of the wider world. For it seems that we are primed to proactively make sense of situations in as much as we organize incoming data in order to fit the theory we have, or are in the process of developing. We expect the world to make a modicum of sense and this expectation helps to catalyze the process of learning. However, there is a trade-off here: Our readiness to find theoretical narratives behind the events around us imputes a propensity to find them even when they are not there. It’s a bit like the faces we see in the clouds – we are innately sensitive to the face configuration and we are therefore inclined to see them in places where there are no faces. The hazard of human theoretical creativity is that it is in constant danger of taking off into the wilder blue yonder and into the realm of fantasy. Garfinkel’s experiment, which I mentioned in the last post, bears this out.

What helps to control the fires of human theoretical creativity and keep it anchored to the real world is that circumstances often force theoretical narratives to be used in an anticipatory way. Although this check is not foolproof, it certainly helps. The child who quickly makes sense of the complex stream of language is rewarded when that learning is confirmed by successful anticipation of the meanings intended by fellow humans. In contrast, however, we note that the patients in Garfinkel’s experiment were not asked to anticipate the therapist’s responses in advance. Reckoning day never came and the patient did not have to put his money where his thoughts were. Each new response was fitted into a growing theoretical elaboration.

There is much to be said for
Gregory Chaitin's view that theories are a form of data compression. That is, theoretical narratives are very succinct expressions that can be used to inform us about the outcome of a large number of experimental cases. If Chaitin is right then Garfinkel’s patients where engaged in an impossible task; that is, of trying to “compress” a random sequence into a succinct narrative explaining the therapists random replies. As each random ‘yes’ or ‘no’ comes in the patient has to incorporate it into his explanation of what is going on, and this requires that the patient, in order to keep track of the irreducible complexities of randomness, must elaborate an explanation that has an equal complexity. As long as the patient is prepared to keep elaborating his explanation of what is going in order to fit an apparently random output, no compression of concept is actually taking place - it is simply a fancy way of recording the data, with no real attempt at radical anticipation of any fundamental underlying logical structure constraining a whole edifice of potential output. The patient was, in effect, acting a bit like one of the conspiracy theorists of our time, who succeed in retrospectively accounting for all sorts of behavior by simply elaborating the conspiracy theory via the addition of further characters, motives and what have you. Seldom does the conspiracy theorist put his theory to real use and launch out with some honest theoretical risk taking by using his theory to generate (that is, anticipate) further incoming data.

To my mind Chaitin is right on at least one count. Right because much of our world, particularly in physics, is organized to yield to those natural philosophers who seek compressed explanatory narratives. However, the fact that Garfinkel’s patients succeeded in retrospectively interpreting random data is testimony to just how complex the socio-personal world is – it is so complex that it can be used to account for random configurations. Although the patients ultimately put together spurious explanatory narratives, the potential is clearly there: The socio-personal world is complex enough for patients to assemble its many elements and variables in order to produce at least a plausible explanation for an ostensibly random output, and this is evidence of just how complex objects are humans and their societies.
Unlike physics the starting point of socio-personal explanations is the most complex things in the cosmos, namely human beings and society. As is often the case when we are dealing with the socio-personal, such as, say, trying to account for simple scene of crime clues, the explanation of the simple is found in the complex purposes of human beings and the condition of society as a whole; in fact we have here an inversion of the physical paradigm: In the latter simple physical principles explain complex physical outputs, but in the social sciences complex socio-personal constructs may be used to account for simple events. In short, Chaitin’s ideas, although certainly very useful, are not comprehensive.

But the socio-personal world is not the only context where complex entities are used as an explanatory framework. In spite of the fact that much of physics yields to Chaitin’s notion that theories are ways of compressing complex physical data, it is precisely in physics that we find a hard case of theoretical incompressibility. We have, of course, the well known random outputs of quantum mechanics, which as far as we yet know are irreducibly random. Although the statistics of quantum mechanics are constrained by various quantum equations, nevertheless particular outputs have a random element and we have to proceed as did Garfinkel’s patients – each real quantum output event is what it is, and has to be taken at face value and simply recorded as part of the complex sequence of random data of which it is part.
Coming Soon: American Antigravity, Tim Ventura and the Philadelphia Experiment.

No comments: