Glossary

Epistemic luck

Also called: veritic luck, anti-luck epistemology

Definition

The way a belief can be true by accident, in a manner incompatible with its being knowledge, even when it is well formed or justified.

Epistemic luck is the way a belief can be true by accident, in a manner incompatible with its being knowledge. That knowledge excludes this kind of luck is one of the few points on which nearly all epistemologists agree; Unger made it the core of an analysis of knowledge, and Pritchard built an anti-luck epistemology around it.

Not all luck undermines knowledge. It may be lucky that I looked up just as a comet passed, but my belief that it passed is not thereby true by accident. Engel distinguished such evidential luck from veritic luck, and Pritchard isolated the kinds that matter: veritic luck, where it is a matter of luck that a belief is true given how it was formed, and reflective luck, where its truth is lucky from the believer’s reflective standpoint. Gettier cases involve veritic luck. Modal conditions such as safety and sensitivity are the main attempts to say what excluding it requires.

The concept matters for AI because frequency-reliable sources produce a distinctive kind of luck. A broken thermometer in a thermostatted room is correct every time it is read, for a stable and explicable reason. A belief formed by reading it is true, and not by a fluke in the ordinary sense. Yet its truth is accidental relative to the temperature, since nothing about the temperature explains what the thermometer displays. Grounded frequency reliability is non-accidental relative to the mechanism producing it while remaining accidental relative to the truth of the particular content.

That is the structure I find in language-model outputs, on the best current evidence about what governs their production. A user who comes to a true belief from such a source is not lucky in the gambler’s sense; the source is right far more often than chance. But if the truth of what it asserted played no role, through its own operation, in why it asserted it, her belief is true by a kind of luck that precludes testimonial knowledge, however impressive the aggregate record. What the record can give her is the warrant of an instrument she has validated.

Questions people also ask

What does epistemic luck mean?

It is the way a belief can turn out true by accident. If you are right only by luck, in the relevant way, you do not know, even if your belief was formed carefully and with good reasons.

What is an example of epistemic luck?

Bertrand Russell’s stopped clock. You look at a clock that stopped exactly twelve hours ago and come to believe the right time. Your belief is true and reasonably formed, but it is true by accident, so it is not knowledge.

What is the difference between veritic and evidential luck?

Evidential luck is luck in coming to have your evidence, as when you happen to look up just as a comet passes. It is compatible with knowledge. Veritic luck is luck in the belief’s being true, given how it was formed, and it is the kind that rules knowledge out. Duncan Pritchard drew the distinction sharply.

Is all luck bad for knowledge?

No. It may be lucky that you were in a position to learn something at all. What knowledge excludes is a belief whose truth, given how you arrived at it, was a matter of luck.

In the library

Further reading

  1. Unger, P. (1968). An analysis of factual knowledge. The Journal of Philosophy 65(6): 157–170.doi.org/10.2307/2024203
  2. Engel, M. (1992). Is epistemic luck compatible with knowledge? The Southern Journal of Philosophy 30(2): 59–75.doi.org/10.1111/j.2041-6962.1992.tb01715.x
  3. Pritchard, D. (2005). Epistemic Luck. Oxford: Oxford University Press.doi.org/10.1093/019928038X.001.0001