I am pleased to see a detailed review of my book, Epistemology and Cognition, in Artificial Intelligence, since the book's principal aim, as Stephen Smoliar indicates, is to build interdisciplinary bridges. Smoliar has made a valiant effort to convey the essence of a long and (arguably) complex book. However, there are a number of misunderstandings I would like to clear up.
First, let me clarify my treatment of justificational rules, logic, and psychology. The concept of justified or rational belief is a core item on the agenda of philosophical epistemology. It is often discussed in terms of "rules" or "principles" of justification, but these have normally been thought of as derivable from deductive and inductive logic, probability theory, or purely autonomous, armchair epistemology. Philosophers have generally assumed that these normative rules can be ascertained without attention to empirical investigation of human cognition. I am trying to "shake the tree of epistemology," as Smoliar puts it. I argue that cognitive science has important contributions to make. The argument proceeds by showing that adequate rules--rules that capture our intuitive understanding of what justified belief is--cannot be derived from pure logic (i.e., logic considered as a set of model-theoretic or proof-theoretic truths). Justification rules must license patterns of reasoning, where reasoning is here understood as transitions from prior belief states to later belief states (or more generally from "cognitive states" to "cognitive states," where prior states may include perceptual states, and output states may include subjective probabilities rather than full-fledged beliefs). But the subject matter of pure logic does not address cognitive states or permissible transitions among them; it just addresses semantic or proof-theoretic properties of sentences. Ultimately I argue that justification rules must authorize psychological processes, which again do not fall within the purview of pure logic.
To illustrate the phases of the argument, let us ask what a logic-generated rule might be like. It might authorize a cognizer to start believing at time t any proposition Q logically implied by his antecedent set of beliefs B. But this is intuitively wrong as a principle of justification. For one thing, it ignores the fact that old beliefs may sometimes need revision, and new inferences from them should be suspended. So it isn't always permissible to infer Q from B. Furthermore, if Q is a remote or obscure consequence of B, the cognizer isn't necessarily justified in starting to believe Q; for he may not grasp or appreciate the relationship between them. Using these and other manoeuvres, I contend that justification is partly a matter of how a cognizer (psychologically) arrives at his beliefs, and hence is a matter of psychological process as well as logic.