We are looking for a theory of knowledge and evidence that confirms and explains the conclusions of Chapters 2 through 6. More specifically, we are looking for a theory that explains (a) why not all evidential relations are inferential; (b) how sensory evidence in particular can be non-inferential (or how beliefs about the world can be evidentially grounded in sensory appearances yet not inferred from sensory appearances); (c) how some knowledge can be foundational (or how some knowledge can be based on evidence which is not itself in need of further justifying reasons); and (d) how inferences that are only contingently reliable can nevertheless give rise to knowledge. Our theory should also explain (e) how knowers can be sensitive to the reliability of their inferences, and even though such inferences are only contingently reliable, so knowers cannot just “see” that they are reliable by a kind of logical intuition into necessary relations.

I have already suggested that agent reliabilism does all of these things. In this chapter I develop and defend that claim. The argument occurs in two stages. In Part I, I argue that reliabilist theories in general confirm all of the conclusions noted in (a) through (d). In other words, “simple” or “generic” reliabilism explains why the skeptical assumptions rejected in Chapters 2 through 6 are false. In Part II, I argue that agent reliabilism is the best version of reliabilism. This is because, in part, agent reliabilism explains (e), or how knowers can be sensitive to their contingent reliability without falling into Humean circularity problems.