Connectionism and the Emergence of Propositional Knowledge

Veikko Rantala

It is argued by many cognitive scientists, and it will be my hypothesis here, that an individual's propositional knowledge, and hence information he tries to transmit in communication by means of symbols, emerges from his nonpropositional knowledge, which is not symbolic but rather representable by means of subsymbolic processes. Therefore, communication breakdowns in intersubjective information which result from variations in meaning are often due to variations in nonpropositional knowledge. It follows that in order to explain such breakdowns and figure out their logical basis, the relation between the propositional and nonpropositional levels of knowledge, and the emergence of the former from the latter, must be better understood. Another relation that is in need of clarification pertains to the question of how the level of (so-called) cultural or public knowledge emerges from the level of individual knowledge. If these relations were better understood, it would also help us to discover the logical principles of different levels of knowledge representation. The nature of these principles has been a controversial question of epistemic logic since the beginning of the sixties.

In this paper, I shall suppose that individual knowledge (whether propositional or nonpropositional) is representable by means of connectionist systems, unlike public knowledge, which can be represented in terms of symbol systems. The former is a controversial assumption, much discussed in the literature, but if it proves to be correct, it may explain why the relation of the two kinds of knowledge is problematic. It can be pointed out, however, that there obtains a limiting case correspondence of a symbolic (Turing) representation to a connectionist one, so that there is a certain conceptual relationship between them. Therefore, the former is kind of conceptually dependent on the latter. There is, furthermore, an approximate correspondence (the nature of which is not, however, very clear at present), which means that the former provides an approximation of the latter. Implications of there existing these two kinds of correspondence for epistemic logic will be preliminarily discussed.