Epistemology
Harman wrote an early discussion of "inference to the best explanation" and argued in later work that all inference or reasoning should be conceived as rational "change in view," balancing conservatism against coherence, where simplicity and explanatory considerations are relevant to positive coherence and where avoiding inconsistency is relevant to negative coherence. He has expressed doubts about appeals to a priori knowledge and has argued that logic and decision theory are theories of implication and consistency and should not be interpreted as theories that can be followed: they are not theories of inference or reasoning.
In Thought and "Change in View" Harman argued that intuitions about knowledge are useful in thinking about inference. More recently, he and Brett Sherman have suggested that knowledge can rest on assumptions that are not themselves known. He and Sanjeev Kulkarni have suggested that elementary statistical learning theory offers a kind of response to the philosophical problem of induction.
Read more about this topic: Gilbert Harman