Epistemic Defeat: Defeaters and Undercutters

Core Idea

A defeater is a belief that undermines or rebuts justification. A rebutting defeater is evidence for the negation of the target (evidence against p); an undercutting defeater is evidence that the justification is unreliable (evidence that one shouldn't trust one's evidence for p). Formally, defeat can be modeled using preference structures or non-monotonic logic: a belief is justified given the available justification unless a defeater is present. Defeaters are crucial for solving Gettier problems.

Explainer

Start from justified true belief: for a belief to count as knowledge, it must be true, believed, and justified. Gettier problems showed that these three conditions aren't sufficient — you can satisfy all three and still fail to have knowledge. The theory of epistemic defeat approaches this problem from a different direction: rather than searching for a fourth positive condition to add, it asks when and how an existing justification can be *undermined*. A defeater is any information that, upon being acquired, reduces or eliminates the justificatory support you had for a belief.

A rebutting defeater provides direct evidence against the target belief — it gives you positive reason to think the belief is false. If you believe the patient has condition X based on their symptoms, and a definitive lab result comes back ruling out X, that result rebuts your belief. The evidence points in the opposite direction from your original conclusion. A rebutting defeater attacks the conclusion itself.

An undercutting defeater works differently and is subtler. It doesn't provide evidence that your belief is false; it provides evidence that your justificatory process is unreliable in this context. Imagine you see a red barn in a field and form the justified belief "that is a red barn." Now you learn that this region is full of cardboard barn facades, painted realistically, that you cannot distinguish from real barns at road distance. This new information doesn't tell you that you're looking at a fake — you might well be looking at the only real barn in the area. But it undermines your ability to trust your perceptual process here: your visual experience of a barn-shape no longer provides strong justification for believing it's a barn, because that same experience would arise whether you were seeing a real barn or a fake. The undercutter attacks the *inference from evidence to belief*, not the belief itself.

The connection to Gettier is direct. In Gettier cases, your justified belief happens to be true, but not because your justificatory process reliably tracked the truth — there's a kind of epistemic luck involved. Defeater theory suggests that a complete account of knowledge must include not just positive conditions but also the absence of undefeated defeaters: your justification must remain intact. Some responses to Gettier formalize this as a no-defeater condition — roughly, there must be no true proposition that, if you learned it, would defeat your justification. This approach captures an important intuition: genuine knowledge is robust to the discovery of new truths, whereas Gettier-style lucky belief is not. The challenge for this account is specifying precisely which defeaters count as relevant and handling cases where different defeaters interact — a defeater can itself be defeated by a further defeater-defeater, generating complex chains of epistemic status that formal models attempt to track.