Abstract

Broadly construed, this dissertation addresses a question central to normative epistemology: "what makes for good thinking?" My answer is a computational, internal, pragmatic, coherence epistemology. I call it, somewhat incompletely and inaccurately, "belief-desire coherence". It is designed to draw from progress in artificial intelligence and cognitive psychology. ;Probably the standard philosophical answer to "what makes for good thinking?" is a variation on "thinking directed toward the truth." I save the bulk of my arguments against this traditional alethic approach for the fourth chapter; in the first three, I motivate my positive, pragmatist alternative. ;Chapter one focuses on the controversial topic of epistemic guidance. I begin by exploring what it is for a creature---natural or artificial---to be intelligent. Assuming that adaptability is fundamental to intelligence, and that learning is fundamental to adaptability, I develop an account of what it is for a creature to learn to think better. Starting with a functional characterization of creatures, I argue that for a creature to learn better thinking requires a feedback mechanism internal to its cognition. The result is a naturalistic and more precise version of "internal" epistemology that captures and explains its basic intuitions. This characterization of internal epistemology suggests, in turn, that the internally available standard for better thinking is pragmatic, to do ultimately with fulfilling the creature's basic aims. ;In the second chapter I consider the wishful thinking objection to any internal pragmatic epistemology, and in response argue for a modified coherentist approach to the evaluation of both beliefs and desires. An internally measurable standard of good thoughts, both desire-like and belief-like, is the level of coherence among them. The proposed coherence has foundational elements in the "default" thoughts that come with the fundamental design of the creature. ;This pragmatic coherence measure, I claim, can provide the internal feedback required for learning. In the third chapter I show how to model this coherence and feedback computationally. Then, with the full theory in place, I outline its several advantages for cognitive science, accounts of folk psychology and emotions, and even ethics