Gettier through Reliabilism and Pramāṇya — Epoche B2
Fake Barns and Defective Causes: Why Reliability Does Not Solve the Gettier Problem The claim After 1963 epistemology divided, roughly, into two repair strategies. One adds a modal condition ruling out luck: the belief must be sensitive, or safe. The other looks at how the belief was produced: it must issue from a reliable process. It is often said that the second strategy is the stronger of the two, and that classical Indian theories of prāmāṇya — the warrant carried by a knowledge-episode in virtue of the means that produced it — point the same way [1] . This essay argues that the second strategy does not solve the Gettier problem either [2] . The fake-barn case shows that a reliable process type and an accidentally true belief can coexist, and the obvious repair — describing the process more narrowly — runs into the generality problem. The Nyāya analysis of pramā is worth putting beside this because it takes the opposite option [3] , individuating at the level of the particular episode and adding an explicit defect clause, and pays the corresponding price. The difficulty is therefore structural rather than a failing of one theory. Virtue epistemologists deny this and claim a third option; the penultimate section takes their claim seriously. Gettier cases and what makes them work The analysis under attack says that S knows that p just in case p is true, S believes p, and S is justified in believing p. Gettier's 'Is Justified True Belief Knowledge [4] ?' (1963) gives two counterexamples of the same shape. Smith has excellent evidence that Jones owns a Ford; he infers, and believes, that either Jones owns a Ford or Brown is in Barcelona. Jones has in fact sold the car. Brown, by coincidence, is in Barcelona. The disjunction is true, Smith is justified in believing it, and Smith does not know it. Two things are doing the work. Justification is inherited from a belief that is false. Truth arrives from a direction the justification knows nothing about. The diagnosis usually offered is epistemic luck : it is an accident, relative to Smith's grounds, that his belief is true. One caution about a familiar illustration: the stopped clock, read at the one moment in the day when it happens to show the right time, is Bertrand Russell's [5] , from Human Knowledge: Its Scope and Limits (1948). It predates Gettier by fifteen years and involves no inference from a falsehood. The causal turn Alvin Goldman's 'A Causal Theory of Knowing' (1967 [6] ) drew the obvious moral: what is missing in Gettier's cases is a causal connection. Smith's belief is not caused, in any suitable way, by Brown's being in Barcelona. Goldman's proposal was that knowledge requires the fact that makes the belief true to be appropriately causally connected with the belief [7] . It handles the cases cleanly, and it fails on knowledge of general truths and of the future, where the fact known cannot cause anything in the knower. Two refinements pursued the same intuition modally. Fred Dretske's 'Conclusive Reasons' (1971 [8] ) holds that R is a conclusive reason for p only if R would not have been the case unless p were. Robert Nozick's Philosophical Explanations (1981 [9] ) develops a tracking account whose central condition is sensitivity : if p were false, S would not believe it. In the stopped-clock case the believer would have believed the same thing had the time been different, so his belief is insensitive and is not knowledge. Precision matters here, because the two modal conditions are constantly swapped. Sensitivity is Nozick's; safety — that S would not easily have believed p falsely — is a different and later condition, due to Ernest Sosa and to Timothy Williamson in Knowledge and Its Limits (2000 [10] ). They are not equivalent. Sensitivity implies that I do not know I am not a brain in a vat, since if I were, I would believe I was not, and it therefore forces the denial of closure under known entailment — a consequence Nozick and Dretske both accepted deliberately. Safety was designed to preserve closure. Process reliabilism Goldman's 'What Is Justified Belief?' (1979) shifts the question from the belief's causal ancestry in the fact to the character of the mechanism that produced it. A belief is justified when produced by a reliable process — one that tends to yield truths — and knowledge is true belief so produced. The theory is externalist: whether the process is reliable is not something the believer needs to establish, or even to know about. This is a genuine advance on the internalist analysis. It does not solve Gettier's problem. The fake barns The decisive case is Goldman's own, published in 'Discrimination and Perceptual Knowledge' (1976) and credited there to Carl Ginet. Henry is driving through the countryside with his son and points out a barn. He is right: it is a barn, and he sees it plainly, in good light, at close range. But unknown to him the district is full of barn façades — convincing fronts with nothing behind them — and had he pointed at any of the others he would have been wrong. Almost everyone judges that Henry does not know there is a barn there. Notice what has happened to reliabilism. The process that produced Henry's belief is ordinary visual perception of a medium-sized object in good conditions, which is about as reliable as human processes get. The belief is true. And still it is a Gettier case, because it is an accident, relative to how the belief was formed, that Henry happened to be looking at the one real barn. Reliability of the type and accidentality of the token have come apart. Whatever excludes Gettier cases, it is not the reliability of the process. The generality problem The natural repair is to describe the process more narrowly: not "vision", but "vision exercised in barn-façade country", which is unreliable. This is the move that runs into what Earl Conee and Richard Feldman set out as the generality problem in 'The Generality Problem for Reliabilism' (1998). Every belief is