Process Reliabilism

Core Idea

Alvin Goldman's process reliabilism holds that a belief is justified if and only if it is produced by a reliable cognitive process — a process that tends to produce true beliefs in the actual world and in nearby possible worlds. Perception, memory, and valid deductive inference count as reliable processes; wishful thinking, astrology, and hasty generalization do not. Reliabilism is an externalist theory: the agent need not be aware that the process is reliable for the belief to count as justified. This approach promises to handle Gettier cases (Gettier beliefs arise from processes that fail to reliably track truth in the relevant case) and explain why children and animals can have knowledge.

How It's Best Learned

Apply reliabilism to a range of cases: perceptual beliefs, clairvoyance, testimony, inference to the best explanation. The generality problem — how to individuate the relevant process type — is the central technical challenge to work through.

Common Misconceptions

• Reliabilism does not say that a reliable process must succeed in every case, only that it has a sufficiently high success rate in the relevant reference class.
• The generality problem is not solved by appealing to common sense; it requires a principled criterion for which process description is the relevant one.

Explainer

From your work on justified true belief, you know the classical account: knowledge is believing something true for good reasons. From Gettier problems, you know that account fails because good reasons can lead to true beliefs by accident. Various responses to Gettier tried to patch the definition — adding a "no false lemmas" clause, requiring causal connections to the fact — but each patch faced new counterexamples. Process reliabilism, developed by Alvin Goldman, steps back and asks a different question: instead of analyzing what makes a *belief* justified, what makes the *cognitive process that produced it* a good one?

Goldman's answer is reliability: a cognitive process is epistemically good if it tends to produce true beliefs. Perception counts — your visual system reliably tracks features of the physical environment. Memory counts — it reliably preserves information about your past experiences. Valid deductive inference counts — if your premises are true and the inference is valid, the conclusion must be true; reliability is maximal. What doesn't count: wishful thinking (wanting something to be true is no guide to whether it is), reading horoscopes, or forming beliefs through pure random guessing. These processes have low truth-producing ratios in the actual world, so beliefs produced by them lack justification.

The critical innovation is externalism: you don't have to know that your belief-forming process is reliable for your belief to be justified. A child who forms accurate perceptual beliefs doesn't need to understand optics or neuroscience; a dog that correctly identifies its owner doesn't need to reflect on the reliability of its nose. Justification is constituted by the actual relationship between the process and truth — it's a fact about the world, not about what the agent can introspect. This contrasts with internalist theories (like classical foundationalism), which require justification to be accessible from the agent's first-person perspective. Internalism struggled to explain how children and animals could count as knowing anything; reliabilism handles them naturally.

The main technical challenge for reliabilism is the generality problem: every token belief is produced by a process that can be described at many levels of generality. My current visual belief was produced by "using my eyes," but also by "using my eyes in good lighting," and also by "using my eyes while sitting in this chair in this building on this day." Which description picks out the relevant process type for evaluating reliability? "Using eyes" is reliable; "using eyes under certain specific conditions on a specific day" might be reliable or unreliable depending on how narrowly we define it. There is no principled answer that Goldman's original formulation provides. This doesn't refute reliabilism, but it means the theory is incomplete until we can say, non-arbitrarily, which process description is the right one to evaluate. Working through this problem is the central task of reliabilist epistemology beyond the introductory level.