Deleuze's unusual metaphysics entails an equally atypical epistemology, or what he calls a transformation of "the image of thought". According to Deleuze, the traditional image of thought, found in philosophers such as Aristotle, Descartes, and Husserl, misconceives of thinking as a mostly unproblematic business. Truth may be hard to discover—it may require a life of pure theorizing, or rigorous computation, or systematic doubt—but thinking is able, at least in principle, to correctly grasp facts, forms, ideas, etc. It may be practically impossible to attain a God's-eye, neutral point of view, but that is the ideal to approximate: a disinterested pursuit that results in a determinate, fixed truth; an orderly extension of common sense. Deleuze rejects this view as papering over the metaphysical flux, instead claiming that genuine thinking is a violent confrontation with reality, an involuntary rupture of established categories. Truth changes what we think; it alters what we think is possible. By setting aside the assumption that thinking has a natural ability to recognize the truth, Deleuze says, we attain a "thought without image", a thought always determined by problems rather than solving them. "All this, however, presupposes codes or axioms which do not result by chance, but which do not have an intrinsic rationality either. It's just like theology: everything about it is quite rational if you accept sin, the immaculate conception, and the incarnation. Reason is always a region carved out of the irrational—not sheltered from the irrational at all, but traversed by it and only defined by a particular kind of relationship among irrational factors. Underneath all reason lies delirium, and drift."
Deleuze's peculiar readings of the history of philosophy stem from this unusual epistemological perspective. To read a philosopher is no longer to aim at finding a single, correct interpretation, but is instead to present a philosopher's attempt to grapple with the problematic nature of reality. "Philosophers introduce new concepts, they explain them, but they don't tell us, not completely anyway, the problems to which those concepts are a response. [...] The history of philosophy, rather than repeating what a philosopher says, has to say what he must have taken for granted, what he didn't say but is nonetheless present in what he did say."
Likewise, rather than seeing philosophy as a timeless pursuit of truth, reason, or universals, Deleuze defines philosophy as the creation of concepts. For Deleuze, concepts are not identity conditions or propositions, but metaphysical constructions that define a range of thinking, such as Plato's ideas, Descartes's cogito, or Kant's doctrine of the faculties. A philosophical concept "posits itself and its object at the same time as it is created." In Deleuze's view, then, philosophy more closely resembles practical or artistic production than it does an adjunct to a definitive scientific description of a pre-existing world (as in the tradition of Locke or Quine).
In his later work (from roughly 1981 onward), Deleuze sharply distinguishes art, philosophy, and science as three distinct disciplines, each analyzing reality in different ways. While philosophy creates concepts, the arts create novel qualitative combinations of sensation and feeling (what Deleuze calls "percepts" and "affects"), and the sciences create quantitative theories based on fixed points of reference such as the speed of light or absolute zero (which Deleuze calls "functives"). According to Deleuze, none of these disciplines enjoy primacy over the others: they are different ways of organizing the metaphysical flux, "separate melodic lines in constant interplay with one another." For example, Deleuze does not treat cinema as an art representing an external reality, but as an ontological practice that creates different ways of organizing movement and time. Philosophy, science, and art are equally, and essentially, creative and practical. Hence, instead of asking traditional questions of identity such as "is it true?" or "what is it?", Deleuze proposes that inquiries should be functional or practical: "what does it do?" or "how does it work?"