Process reliabilism (hereafter reliabilism) is a theory of epistemic justification that claims that a belief’s justificatory status is determined by the reliability of the process that formed the belief. Here, a process is a functional operation that maps inputs to outputs (ie. input: experiential state; output: belief), and the processes we are concerned with are cognitive processes. A process is reliable if and only if it would yield a high proportion of truths over a wide variety of situations of the type we typically encounter (Alston). Beliefs that are the product of reliable belief-forming processes are justified, and justified to the extent that the process is reliable, while those beliefs that are not are unjustified.
Goldman gives the following reliabilist account of justification:
1. If S’s belief that p at t results (immediately) from a belief-independent process that is (unconditionally) reliable, then S’s belief that p at t is justified.
2. If S’s belief that p at t results (immediately) from a belief-dependent process that I (at least) conditionally reliable, and if the beliefs (if any) on which this process operates in producing S’s belief that p at t are themselves justified, then S’s belief in p at t is justified.
A belief-independent process is a process that does not have a belief state as an input. A belief-dependent process has at least one belief state as an input. A process is conditionally reliable when a sufficient percentage of output beliefs are true given that the input beliefs are true.
Goldman complicates this account somewhat in trying to account for defeaters. He claims that the justificatory status of a belief is also a function of the processes that could or should have been employed. Thus, for a belief to be justified it must also be that there was no reliable or conditionally reliable process available to the agent which had it been used in addition to the process actually used would have had the result that S did not believe that p at t.
At first glance, reliabilism is promising. For many justified beliefs one is able to identify a reliable process type that brought it about, and for most unjustified beliefs one is able to identify an unreliable process type that brought it about. Although this account of epistemic justification has some initial intuitive appeal, it faces substantial criticisms. Perhaps the biggest problem facing reliabilism is the Generality Problem. Any particular belief-forming process is a token of an indefinite number of types, and these types often differ radically in their reliability. The problem is that in order to evaluate the claim of reliabilism we need an account of which process type(s) is the relevant type – the type which determines the justificatory status of the belief.
For example, my belief that there is a computer in front of me is the result of a process of all of the following types: visual processes, done on Wednesday processes, processes that lead to true beliefs, processes done at 4:30pm, processes done by a Canadian. Understood as the result of some of these processes, my belief is unjustified; yet understood with other process types as being relevant, my belief is justified. The challenge for the reliabilist is to give a principled account that tells us what the relevant process types are. This task is difficult since if the relevant process types are construed too narrowly, there will only be one token of the relevant type. As such, all true beliefs would be justified and all false beliefs would be unjustified (since reliability would be determined on the basis of only one case). But this is clearly not the case. There are justified false beliefs and unjustified true beliefs. On the other hand, if the relevant process type is construed too broadly, then too many beliefs will inherit the same justificatory status, though the resulting beliefs seem to be of differing statuses. For instance, if perception were a relevant process type, then all beliefs that resulted from perception would be equally justified, but clearly this is not the case. Perception leads to better justified beliefs when the cognizer is in good lighting and looking at medium sized objects (in contrast to tiny objects in the dark).
Alston responds to this problem by defending the claim that for each belief forming process there is only one process type that is ‘psychologically real’, and it is this process type that is relevant for assessing justification. The process that is psychologically real is the one which corresponds to the actual inputs and outputs of the belief forming process. If we are concerned with my belief that there is a computer in front of me, this is the actual output, and the input was aspects of my particular perceptual experience. Alston’s claim is that there is only one function that maps this input to this output – that function is the relevant belief forming process type.
Although this response limits the possibilities for which process type is relevant, Alston has failed in reducing the list to one candidate. There are multiple operative psychological types since the operational function can broadened by broadening the input and outputs (without changing what the inputs and outputs are). The output that I believe there is a computer in front of me is also the output of a function that takes the same input and whose output is either this belief or my belief that there is a typewriter in front of me. Both the input and the output of a function can be broadened via disjunction (among other ways) and this allows for more and more functions (process types) to be active in a case of belief formation. We still need a way of determining which of these ‘active’ functions or processes is the relevant one – the one by which we assess the resultant beliefs justificatory status. Until we have such an account, reliabilism cannot get off the ground because we cannot see what it says about various cases.