This paper discusses two versions of reliabilism: modal and probabilistic reliabilism. Modal reliabilism faces the problem of the missing closeness metric for possible worlds while probalistic reliabilism faces the problem of the relevant reference class. Despite the severity of these problems, reliabilism is still very plausible (also for independent reasons). I propose to stick with reliabilism, propose a contextualist (or, alternatively, harmlessly relativist) solution to the above problems and suggest that probabilistic reliabilism has the advantage over modal reliabilism.

Epistemic luck has been the focus of much discussion recently. Perhaps the most general knowledge-precluding type is veritic luck, where a belief is true but might easily have been false. Veritic luck has two sources, and so eliminating it requires two distinct conditions for a theory of knowledge. I argue that, when one sets out those conditions properly, a solution to the generality problem for reliabilism emerges.

According to reliabilists about epistemic justification, what makes a belief epistemically justified is that it was produced by a reliable process of belief-formation. Earl Conee and Richard Feldman have forcefully presented a problem for such reliabilism, "the generality problem."? The generality problem arises once we realize that the notion of reliability applies straightforwardly only to types of process--for only types of process are repeatable entities which can produce true or false beliefs in each of their instances. Moreover, any token process (...) will be an instance of indefinitely many types of process. Which of these types must be reliable for my belief to be justified, according to reliabilism? That question, generalized to cover every case of belief-formation, is the generality problem for reliabilism. In this paper I propose a solution to the generality problem. The solution makes use of the basing relation, and so, given that it isn't clear how to characterize that relation, it might be thought to replace one problem with another. I argue that, however difficult it is to characterize the basing relation, every adequate epistemological theory must make use of it implicitly or explicitly. Therefore, it is perfectly legitimate to appeal to the basing relation in solving a problem for an epistemological theory. (shrink)

In “Why the generality problem is everybody’s problem,” Michael Bishop argues that every theory of justification needs a solution to the generality problem. He contends that a solution is needed in order for any theory to be used in giving an acceptable account of the justificatory status of beliefs in certain examples. In response, first I will describe the generality problem that is specific to process reliabilism and two other sorts of problems that are essentially the same. Then I will (...) argue that the examples that Bishop presents pose no such problem for some theories. I will illustrate the exempt theories by describing how an evidentialist view can account for the justification in the examples without having any similar problem. It will be clear that other views about justification are likewise unaffected by anything like the generality problem. (shrink)

Guided by the work of William Alston, Jonathan Adler and Michael Levin propose a solution to the generality problem for reliabilism. In some respects their proposal improves on those we have discussed. We argue that the problem remains unsolved.

Critics of reliability theories of epistemic justificationoften claim that the `generality problem' is an insurmountabledifficulty for such theories. The generality problem is theproblem of specifying the level of generality at which abelief-forming process is to be described for the purposeof assessing its reliability. This problem is not asintractable as it seems. There are illuminating solutionsto analogous problems in the ethics literature. Reliabilistsought to attend to utilitarian approaches to choices betweeninfinite utility streams; they also ought to attend towelfarist approaches to social (...) choice situations that donot demand full aggregation of individual welfares.These analogies suggest that the traditional `single number'approach to reliability is misguided. I argue that a newapproach – the `vector reliability' approach – is preferable.Vector reliability theories associate target beliefs withreliability vectors – that is, structured collections ofreliability numbers – and construct criteria of epistemicjustification that appeal to these vectors. The bulk of thetheoretical labor involved in a reliability account of epistemicjustification is thus transferred from picking a uniquereliability number to constructing a plausible criterionof epistemic justification. (shrink)