The library of the mind

In a famous essay Borges wrote of an infinite library that contained all possible books (and most of it nonsense at that). The mind is not like that. It has only a few books in it.

In the philosophy of the cognitive sciences, there are competing views of the nature of the mind. One school, the evolutionary psychologists, hold that the mind is composed of a large number of special purpose modules, each designed by natural selection to do one thing well (enough) and no more. Another school, represented by Jerry Fodor, holds that the central part of the mind (excepting the sensorimotor parts of the central nervous system) is composed of a general inference machine, in which all information is available to each inference, and in which the only rules are those of logic.

I want to comment on a third possibility.

In an essay published in 1998, Richard Samuels proposed what he called the Library Model of Cognition. Instead of the massive number of modules, each shallow, obligate in its behaviour, and shaped by natural selection for one and only one task, Samuels proposed that the mind includes a number of inborn items of knowledge, rather than cognitive mechanisms:

Evolutionary psychologists often talk as if any theory which posits large amounts of innate, domain-specific cognitive structure is a version of MMH [Massive Modularity Hypothesis]. This, however, is not the case. Computational mechanisms are not the only possible kind of innate, domain-specific psychological structure. Another possibility is that humans possess innate, domain-specific bodies of knowledge.

These items are, he claims, sufficiently the result of selection to resolve the problems with the General Cognitive view of Fodor, which is undercut by the problem of knowing enough to be able to solve problems like learning languages (Fodor agrees with Chomsky's Language Acquisition Device view, that resolves the problem of underdetermination or "poverty of stimulus", but it's a good example), without the "Swiss Army Knife" view of modularity for ordinary cognition held by Evolutionary Psychology.

This set me wondering: how would one be able to employ such a Library? What preconditions would there need to be? And what does the Library contain? Not every possible book, of course, because the mind is limited. But sufficient mental representations to get general cognition underway as the child develops and interacts with its environment.

So, suppose that we have a set number of these mental representations at birth, shaped by selection, to help us identify, for example, persistence of objects, which are not domain specific. This supposes that we can employ these representations whenever we need to. They are not static things, but are truth-evaluable, and are tested in each implementation of the Library.

But how do we implement them? When do we know how to employ an item of knowledge, innate or otherwise? At maturity we must have some scheme, but what about when we are little? It seems to me that the library has to include not only the items of knowledge, but also some "access rules", or else innate knowledge does no work.

The solution is, I think, to assume just that - our Library includes heuristics, which bias the ways in which we employ our knowledge base (and these persist once we start to acquire personal items of experiential knowledge too, see Gigerenzer and Todd's work). In short, the Library has access rules up on the wall next to the items on the shelf.

This allows us to have limited rules and items without the need for them to be modular. I have never been that impressed by the argument of the EPists that natural selection favours modularity. It seems to me a fallacy of affirming the consequent. Natural selection modifies things that are variable, yes, but we only have reason to think they are variable once selection has already modified them. There is no a priori reason for thinking that there is something special about modules, except that they are sensitive to selection. They may be independent of other things, or not, in the development of the organism.

And the MMH is overstringent anyway - why should our innate dispositions be informationally encapsulated? There is no need for our Library items to be isolated from the rest of the Library and the uses therefof except that there is supposed to be a performance deficit. And this doesn't follow - consider how sensorimotor skills can be trained, say in a martial artist, to act rapidly and without thought. I'm pretty sure that an ippon-seoi-nage throw is not a module, but for a few minutes in my judo training, I was able to do it without thinking, and rapidly, in sparring. Why not in cognition?

And this also explains why I am a bibliphile. I was born to like libraries...