Gettier Problem
The Gettier problem is the demonstration that justified true belief (JTB) is not sufficient for knowledge, set out by Edmund Gettier in a 1963 three-page paper (see the SEP entry). You can satisfy all three conditions and still intuitively fail to know.
Why is a three-page paper a big deal?
If 2000 years of defining knowledge as JTB collapses on a three-page counterexample, the whole reductive project is suspect.
The traditional JTB analysis: S knows that P iff
- P is true
- S believes that P
- S is justified in believing P
Plato flirted with this in the Theaetetus; modern epistemology ran with it until 1963.
Gettier’s two counterexamples
Smith & Jones (the "Ford or Barcelona" case)
Smith has strong evidence that Jones owns a Ford (he’s seen him drive one for years). From this he infers the disjunction: Either Jones owns a Ford, or Brown is in Barcelona, where Brown’s location is picked arbitrarily. It turns out Jones is actually driving a rental (no Ford), but by coincidence Brown really is in Barcelona.
Smith’s belief is true (by the Barcelona disjunct he didn’t know about), believed, and justified (inferred from well-supported premises). But did Smith know? Plainly not, he got lucky.
The coins case has the same structure: Smith infers a true conclusion through a false intermediate premise.
What Gettier cases show is that truth and justification can come apart by epistemic luck. A belief’s truth can depend on a happy accident rather than proper grounding. Something more than JTB is needed.
The post-Gettier menu (per SEP), each adding or replacing condition 3:
- No-false-lemmas (Clark): the belief can’t rest on any false premise. Fails: some Gettier cases (fake barns) don’t involve a false premise
- Defeasibility: no true proposition would, if added to your evidence, undermine justification. Hard to state non-circularly
- Causal theory (Goldman): the fact must cause the belief. Fails on fake barns: the real barn causes Henry’s belief, but he still doesn’t know
- Reliabilism: the belief must come from a reliable process. Still Gettierable in hostile environments
- Sensitivity (Nozick): if P were false, S wouldn’t believe it. Produces the “abominable conjunction” (you know you have hands but don’t know you’re not a brain-in-a-vat)
- Safety (Sosa, Williamson): in all nearby worlds where S believes P, P is true. Handles fake barns but hangs on which worlds count as nearby
- Virtue epistemology (Sosa, Greco): knowledge is apt belief, true belief that manifests intellectual competence
- Knowledge-first (Williamson): give up. Knowledge is primitive; JTB-plus was misconceived from the start
Fake barns (Ginet/Goldman)
Henry drives through barn-facade country and happens to look at the one real barn. Justified, true, reliably formed, still not knowledge, because truth rides on lucky placement.
My take
What Gettier actually teaches me, practically: “justified + true” is cheap; “justified + true + not-by-luck” is the thing we actually care about, and the post-1963 literature is 60 years of trying to pin down what “not-by-luck” means without smuggling “knowledge” back into the definiens.
I find the safety condition the most intuitive. It says a belief is knowledge when it would still have been true across small perturbations of how you formed it. That matches how I already think about robustness in engineering: a result that works under these exact conditions but breaks one step away isn’t a result you know, it’s a result you got.
The knowledge-first response (Williamson) is the honest one if the whole program is failing. Maybe “knowledge” isn’t built out of simpler pieces, the same way “cause” maybe isn’t built out of constant conjunction. The reductive project was the mistake, not the specific reductions.
Either way, JTB is dead, and any time someone says “well, I’m justified and it’s true, so I know it,” they’re skipping the part where the truth might have nothing to do with their justification.