5.3 Gettier and Other Complications

What this lecture covers

This lecture presents the Gettier problem, the famous set of counterexamples that undermined the justified-true-belief analysis of knowledge introduced in the previous lecture. Millican works through classic Gettier cases, surveys failed attempts to patch the analysis, and closes with the lottery paradox and contextualist worries about whether "knowledge" names a single, unified concept at all.

You'll come away able to construct a Gettier-style counterexample, explain why probability-based patches run into the lottery paradox, and describe how context can shift what counts as "knowing" something in ordinary speech.

Key ideas

• The Gettier structure: if you're justified in believing P, and P clearly implies Q, you seem justified in believing Q—but Q can turn out true only by accident, making the result a justified true belief that isn't knowledge.
• The mirage/oasis case: seeing a mirage justifies believing there's an oasis ahead; by coincidence a real oasis is hidden behind a dune, so the belief is true and justified but not knowledge.
• The job applicant case: justifiably believing a specific well-qualified man will get a job leads to the true belief that "a man will get the job," but only because a different man got it through unrelated circumstances.
• The 'no false premises' patch: adding a fourth condition that justification must not depend on any falsehood—shown to fail with cases like a slightly inaccurate headcount that still supports a true, apparently knowledgeable inference.
• The lottery paradox: no matter how high a probability threshold you set for knowledge, a large enough lottery generates beliefs (each ticket won't win) that meet the threshold yet clearly aren't knowledge, since one ticket must win.
• The accident-avoidance idea and its difficulty: trying to define knowledge as truth that isn't "accidental" is intuitive but hard to pin down precisely, as shown by the corroded-speedometer example.
• Contextualism: how demanding a standard we apply for "knowing" something can shift with the stakes, suggesting the threshold for knowledge may not be fixed.
• Ordinary uses of "know": everyday questions like "does she know he's cheating?" often really ask about belief or about the facts, not about a strict epistemological state, raising a Wittgensteinian doubt about whether "knowledge" has one single essence.

Walkthrough

The Gettier structure and two examples (0:09)

Millican sets up the general pattern behind Gettier cases: justified belief in P, together with a clear logical implication to Q, seems to justify Q as well. He illustrates with the mirage/oasis case and the classic job-applicant case, showing in both that the resulting belief is true and justified by the standard analysis, yet intuitively not knowledge, because the truth is essentially accidental relative to the justification.

Attempts to patch the analysis (3:13)

He describes the historical effort to add a fourth condition—that justification must not rely on any false premise—and shows this fails using the headcount example, where inferring "more than 40 people" from a slightly wrong reported figure of 78 (actually 77) still seems like knowledge despite depending on a falsehood.

The lottery paradox (6:14)

Millican explains why replacing justification with a high-probability requirement doesn't work either: in a billion-ticket lottery, the belief that any single ticket won't win is extremely probable yet clearly not knowledge, since exactly one ticket will win. Since this holds for arbitrarily large lotteries, no probability threshold below certainty can capture knowledge.

Ruling out accidents (8:16)

Considering the idea that knowledge requires ruling out accidental truth, Millican uses the example of a corroded speedometer that happens to still keep a driver within a safe margin, showing how hard it is to say precisely when a true belief counts as "accidentally" true versus reliably grounded.

Contextualism and everyday uses of "know" (10:17)

The lecture closes by examining how the bar for "knowing" something shifts with the importance of the situation, using the example of checking a train's departure time under increasing pressure. Millican also distinguishes everyday uses of "know" that really ask about someone's belief (as in soap-opera dialogue) from uses that really ask about the facts, suggesting—drawing on the later Wittgenstein—that "knowledge" may not have a single unified essence but instead varies with our practical purposes.

Before you watch

• Watch the previous lecture on the traditional justified-true-belief analysis of knowledge, since this lecture is built entirely around counterexamples to that analysis.

Check your understanding

1. What is the general structure that Gettier cases share, and how does the oasis/mirage example fit it?
2. Why does adding a "no false premises" condition fail to rescue the justified-true-belief analysis?
3. How does the lottery paradox show that a high-probability requirement cannot fully capture knowledge?
4. What does the corroded-speedometer example reveal about the difficulty of defining knowledge as "non-accidental" true belief?
5. How does contextualism suggest that the standard for "knowing" something can shift depending on the situation?