The Replacement of General-Purpose Learning Models with Adaptively Specialized Learning Modules

Theories of learning are and always have been predominantly associative theories. However, in the study of animal learning, where these theories have historically been most dominant, a different conception is gaining ground. Whereas associative theories have their historical roots in the empiricist philosophy of mind, the alternative conception has its roots in evolutionary biology, more particularly in zoology, that is, in the study of the natural history of animal behavior and of the mechanisms that enable animals to cope with the challenges posed by their habits of life. Associative theories of learning assume a basic learning mechanism, or, in any event, a modest number of learning mechanisms. These mechanisms are distinguished by their properties--for example, whether or not they depend on temporal pairing--not by the particular kind of problem their special structure enables them to solve. Indeed, people doing neural net modeling, which is currently the most widespread form of associative theorizing, are often at pains to point out that the network has solved a problem in the absence of an initial structure tailored to the solution of that problem (e.g. Becker & Hinton, 1992). The alternative conceptualization, by contrast, takes for granted that biological mechanisms are hierarchically nested adaptive specializations, each mechanism constituting a particular solution to a particular problem. The foliated structure of the lung reflects its role as the organ of gas exchange, and so does the specialized structure of the tissue that lines it. The structure of the hemoglobin molecule reflects its function as an oxygen carrier. The structure of the rhodopsin molecule reflects its function as a photonactivated enzyme. One cannot use a hemoglobin molecule as the first stage in light transduction and one cannot use a rhodopsin molecule as an oxygen carier, any more than one can see with an ear or hear with an eye. Adaptive specialization of mechanism is so ubiquituous and so obvious in biology, at every level of analyis, and for every kind of function, that no one thnks it necessary to call attention to it as a general principle about biological mechanisms. In this light, it is odd but true that most past and contemporary theorizing about learning does not assume that learning mechanisms are adaptively specialized for the solution of particular kinds of problems. Most theorizing assumes that there is a general purpose learning process in the brain, a process adapted only to solving the problem of learning. Needless to say, there is never an attempt to formalize what exactly that