Reframing Epistemic Luck: Agent Reliabilism and the Acquisition of Understanding — Epoche C1
Epistemic luck: the problem stated precisely Epistemic luck is the phenomenon of getting something right by accident rather than by competence, and the reason it dominates modern epistemology is that Edmund Gettier showed in a three-page paper of 1963 that a belief can be true, and justified, and still be right only by accident. His recipe is simple. Take a belief that is well supported by the evidence but false — you have excellent reason to think your colleague Jones owns a Ford, though in fact he does not. Deduce from it something that follows validly: that either Jones owns a Ford or your friend Brown is in Barcelona. Now suppose that, by coincidence, Brown really is in Barcelona. Your conclusion is true; your reasoning was impeccable; you had justification. But nobody wants to say you knew, because the route to the truth passed through a falsehood and the arrival was a fluke. Whatever knowledge requires beyond justified true belief, it requires something that rules that fluke out. The next question is what luck itself amounts to, and here the standard treatment is Duncan Pritchard's. In Epistemic Luck (2005) he analyses luck modally — that is, in terms of what happens in other possible ways the world might easily have gone. An event is lucky, on his account, if it occurs in the actual world but fails to occur in a wide class of the nearest possible worlds in which the relevant conditions for it are the same. Winning a fair lottery is lucky because in almost all the nearby worlds in which you buy that ticket you lose. Pritchard then isolates the species that threatens knowledge, which he calls veritic luck: it is a matter of luck that the belief the agent formed is true, in the sense that in most nearby worlds where the agent forms that belief by that method, the belief is false. The corresponding requirement, stated positively, is safety: a belief is safe when it could not easily have been false while formed in the same way. Gettier's Barcelona believer fails safety flagrantly. Almost any small change to where Brown had gone leaves the belief false, and the believer forming it just the same. One further distinction, drawn by Pritchard in later work, will carry most of the weight of this essay, so it is worth setting out with its example. Alvin Goldman's 1976 paper on perceptual knowledge reports a case devised by Carl Ginet: Henry drives through the countryside, sees a barn, and believes there is a barn before him. He is right, and his eyesight is excellent. What he does not know is that the district is full of barn façades — painted fronts with nothing behind them — and he happened to stop at the one real barn. Nearly everyone judges that Henry does not know. But notice how his case differs from Gettier's. Nothing has intervened between Henry's perceptual competence and the truth; his belief was produced by good vision operating properly on a genuine barn. The trouble is entirely in the neighbourhood. Pritchard therefore separates intervening luck, where something comes between the agent's competence and the fact (the Barcelona case), from environmental luck, where the competence delivers the truth in the ordinary way but the surroundings are treacherous (the barn case). The distinction is not a refinement for its own sake. It is where the theory this essay defends succeeds and where it fails. The virtue-theoretic answer, and a label that needs correcting The response to Gettier that this essay is concerned with locates the missing ingredient in the agent rather than in the belief, and its clearest statement is Ernest Sosa's. Sosa models cognitive performance on an archer's shot, which admits three independent assessments: it may be accurate, meaning it hits the target; adroit, meaning it manifests the archer's skill; and apt, meaning it is accurate because adroit. A shot blown to the target by a gust of wind is accurate but not apt. The proposal is that a true belief counts as knowledge when it is apt — true because of the exercise of the believer's cognitive competence — and the diagnosis of Gettier cases falls out immediately: the Barcelona believer's belief is accurate and adroit but not apt, because the explanation of its truth runs through Brown's travel arrangements rather than through the believer's reasoning. John Greco develops the same thought in terms of credit: knowledge is a cognitive achievement, and an achievement is one whose success is creditable to the agent, meaning that the agent's ability is the most salient element in the causal explanation of the success. Here a correction is due, because the compressed version of this essay carried an attribution that is not standard and that obscures a real difference. 'Agent reliabilism' is Greco's term for Greco's position: the view that what matters epistemically is the reliability of the agent — their stable dispositions and abilities — rather than the reliability of a belief-forming process considered in isolation. Linda Zagzebski's view in Virtues of the Mind (1996) is agent-centred in the same broad way but is normally classified separately, as virtue responsibilism, and the difference matters. For Greco and Sosa, the relevant competences may be brute faculties such as eyesight, which are no credit to anyone's character. For Zagzebski, an intellectual virtue is an acquired excellence of character with two components that must both be present: a motivational component, the agent's motivation to attain cognitive contact with reality, and a reliable success component, the fact that acting on that motivation does in general get one to the truth. Her definition of knowledge is a state of true belief arising out of acts of intellectual virtue, and the phrase 'arising out of' carries the same weight as Sosa's 'because'. So the essay's thesis is best stated across the family: agent-centred virtue epistemology, in either the reliabilist or the responsibilist version, makes epistemic standing depend on whether the truth is at