Justified True Belief and the Gettier Problem

For centuries, philosophers often analyzed knowledge as:

justified true belief.

Then, in 1963, Edmund Gettier published a very short paper.

Its central lesson was devastatingly simple:

A person can have a belief that is:

• justified,
• true

and still not possess knowledge.

The JTB Model

The classical analysis says:

S knows P if and only if:

1. S believes P.
2. P is true.
3. S is justified in believing P.

This seems plausible.

Why Each Condition Matters

Belief:

you must accept P.

Truth:

P must actually be the case.

Justification:

you need good epistemic grounds.

What could be missing?

Gettier’s Insight

The missing ingredient is revealed by epistemic luck.

A justified belief can become true accidentally.

The justification points in the wrong direction, while coincidence makes the conclusion true.

Classic Structure

Many Gettier cases have this form:

1. You have good evidence for false proposition A.
2. You validly infer broader proposition B.
3. B happens to be true for an unrelated reason.

You now have:

• justified belief,
• true belief.

But intuitively:

not knowledge.

Smith and Jones

Gettier’s original examples involve two people often called:

Smith

and:

Jones.

One famous case concerns:

• a job,
• coins in a pocket.

Job and Coins Case

Suppose Smith has strong evidence that:

Jones will get the job.

Smith also has evidence that:

Jones has ten coins in his pocket.

Smith infers:

“The person who will get the job has ten coins in his pocket.”

The Twist

Unknown to Smith:

Smith himself will get the job.

And, also unknown to Smith:

Smith has ten coins in his pocket.

So Smith’s proposition is true.

Does Smith Know?

Smith believes:

the person who gets the job has ten coins.

The proposition is true.

His inference was reasonable.

Yet he does not seem to know it.

The truth arrived by luck.

What Went Wrong?

Smith’s justification depended on:

Jones gets the job.

That premise was false.

The final proposition became true through an unrelated coincidence.

The justification and truth were accidentally connected.

False Lemma Diagnosis

One early response said:

knowledge cannot depend on a false lemma.

Add a fourth condition:

the justification must contain no false premise.

This blocks some Gettier cases.

No False Lemmas

The proposal becomes:

knowledge = justified true belief + no false lemmas.

But new counterexamples can avoid explicit false premises.

The problem survives.

Fake Barn Country

Imagine you drive through an area filled with realistic barn facades.

Only one structure is a real barn.

By chance, you look directly at the real one.

You think:

“That is a barn.”

Your Belief

The belief is:

• true,
• perceptually justified.

Yet many philosophers hesitate to call it knowledge.

You could easily have looked at a facade.

What Makes It Lucky?

Your perceptual method is not safe in the local environment.

You happened to succeed.

Knowledge seems to require more robust connection to truth.

Epistemic Luck

Gettier cases are not ordinary luck.

They involve a specific form:

the belief is true despite a defect in how the truth was reached.

This is epistemic luck.

Environmental Luck

In fake-barn cases, the immediate perception works correctly.

The problem lies in the environment.

Truth is fragile nearby.

This is environmental epistemic luck.

Intervening Luck

In classic Gettier cases, an error occurs and another accident repairs it.

Falsehood intervenes before truth is restored.

This is a different structure.

Safety Condition

One major proposal says knowledge requires safety.

A belief is safe if:

in nearby possible worlds where you form the belief similarly, it remains true.

Fake Barn and Safety

In nearby situations, you look at a facade and believe:

that is a barn.

The belief would be false.

Therefore your actual true belief is unsafe.

Sensitivity

Another condition is sensitivity:

If P were false, would you still believe P?

If yes, your belief may fail as knowledge.

Nozick

Robert Nozick developed a tracking theory of knowledge using sensitivity-like counterfactual conditions.

Knowledge tracks truth across possible situations.

Sensitivity Problem

Some ordinary knowledge appears insensitive.

Example:

I know I am not a brain in a vat.

If I were one, perhaps I would still believe I am not.

Sensitivity can generate skeptical consequences.

Reliabilism

Reliabilists respond by focusing on process reliability.

A belief produced by a sufficiently reliable method may count as knowledge.

This bypasses some internal justification problems.

Gettier and Reliabilism

If the method succeeds only by accident in the relevant environment, it may not be reliable enough.

Fake-barn perception becomes problematic.

Virtue Epistemology

Another approach says knowledge is true belief because of epistemic competence.

The belief succeeds through the agent’s intellectual ability.

Archery Analogy

Imagine an archer hits the target because of skill.

That is achievement.

If a gust of wind knocks a bad shot onto the bullseye, success is lucky.

Knowledge should resemble success through competence.

Ernest Sosa

Ernest Sosa developed influential virtue-epistemological accounts.

He compares beliefs to performances that can be:

• accurate,
• adroit,
• apt.

Apt Belief

A belief is apt when it is true because it was competently formed.

This directly targets Gettier luck.

Animal Knowledge

Sosa sometimes distinguishes:

• animal knowledge,
• reflective knowledge.

Animal knowledge can arise from reliable competence without explicit meta-justification.

Reflective knowledge includes awareness of competence.

Credit Theory

Knowledge may require that truth be creditable to the knower’s ability.

The agent deserves epistemic credit for getting it right.

Knowledge-First Epistemology

Timothy Williamson challenges the attempt to analyze knowledge into simpler components.

He treats knowledge as fundamental.

Instead of defining knowledge through belief + extras, start with knowledge itself.

Why Knowledge-First?

Gettier cases suggest every proposed decomposition can be patched and counterexampled.

Perhaps knowledge is not reducible to a checklist.

Williamson’s View

On a knowledge-first approach:

knowledge is a basic mental state.

Belief may sometimes be analyzed in relation to knowledge rather than the reverse.

This reverses traditional methodology.

Causal Theory of Knowledge

Another historical proposal says the fact known must appropriately cause the belief.

If you know:

there is a tree,

the tree should causally contribute to your belief.

Causal Connection

This works well for perception.

But what about:

• mathematics,
• future facts,
• universal laws?

Causal theories have limited scope.

Proper Function

Alvin Plantinga proposed epistemic warrant connected to cognitive faculties functioning properly in suitable environments.

Knowledge requires more than subjective justification.

This adds externalist structure.

Defeasibility Theory

A belief may count as knowledge only if no true defeating information exists.

Gettier cases often contain hidden defeaters.

Hidden Defeater

Smith’s belief is defeated by:

Jones will not get the job.

Smith does not know this.

But the hidden fact undermines his justification.

Trouble with Defeaters

Defining exactly which defeaters matter is difficult.

Too broad a rule creates skepticism.

Too narrow a rule lets Gettier cases through.

The General Pattern

Every proposed fourth condition faces pressure:

• no false lemmas,
• safety,
• sensitivity,
• reliability,
• competence.

Each captures something important.

None has produced universal consensus.

Why Gettier Matters

Gettier did not merely find weird edge cases.

He revealed that:

truth + justification

must connect in the right way.

Knowledge is an achievement, not an accidental intersection.

Truth-Connection

A good theory of knowledge must explain:

why the belief is true.

The justification should be appropriately related to the fact.

This is the core insight.

Evidence Can Mislead

Smith has good evidence.

Yet reality differs.

This shows:

justification is perspective-relative.

Truth is world-relative.

Knowledge must bridge the two.

Internalist Tension

Internalists emphasize what reasons the person can access.

Gettier shows internally excellent justification can still be unlucky.

This motivates externalist conditions.

Externalist Tension

Externalists add:

reliability, safety.

But then a person may know without knowing why they know.

This can feel unsatisfactory.

Epistemic Luck Cannot Be Eliminated Completely

Every belief involves some luck.

You did not choose:

• reliable eyesight,
• favorable environment.

The goal is not zero luck.

It is eliminating the wrong kind.

Modal Robustness

Knowledge should remain true across relevant nearby possibilities.

This is why modal concepts such as:

• safety,
• sensitivity

became important.

Knowledge as Achievement

Another intuitive summary:

Knowledge is true belief attributable to epistemic competence rather than accident.

This captures much of the post-Gettier landscape.

Science and Gettier

Scientific conclusions can also be accidentally true.

A flawed experiment may yield the correct result for the wrong reason.

That is not strong scientific knowledge.

Replication and mechanism reduce epistemic luck.

Measurement Error

Suppose a broken instrument always reads:

100.

You measure a sample that happens to be exactly 100.

The reading is true.

But you do not know the value through that instrument.

This is a practical Gettier-like case.

Broken Clock

A stopped clock reads:

3:00.

You look at exactly 3:00.

You form the true belief:

it is 3:00.

Do you know?

Normally:

no.

The method is unreliable.

Why Broken Clock Is Different

The belief is true by coincidence.

The source does not track time.

This cleanly illustrates reliability.

Gettier and AI

An AI system can produce a correct answer for the wrong reasons.

Example:

it exploits a spurious pattern.

Correct output does not guarantee genuine knowledge.

Shortcut Learning

A model may classify wolves by:

snow in background.

It gets many answers right.

But the representation tracks the wrong feature.

This resembles epistemic luck.

Robust Generalization

A stronger AI knower should succeed across changed contexts because it tracks the relevant structure.

Robustness is an epistemic virtue.

Explanation

If a system can explain:

why P is true

using the right causal structure, confidence in knowledge increases.

But explanation can itself be post hoc.

Mechanistic validation still matters.

Social Epistemology

Institutions reduce Gettier-like luck through:

• peer review,
• replication,
• calibration.

Knowledge is often made robust collectively.

Gettier and Certainty

The problem is not lack of certainty.

Smith may feel very confident.

The issue is structural:

his truth is accidental.

Confidence cannot fix that.

Knowledge Is Not Justified True Belief

This is the enduring conclusion.

JTB captures necessary ingredients.

It does not capture their correct connection.

The Philosophical Lesson

Gettier showed that knowledge requires more than:

• belief,
• truth,
• justification.

The belief must be true because of the right epistemic relationship to reality.

Exactly how to define that relationship remains contested.

The Next Question

If knowledge requires justification, we must ask:

What counts as good justification?

How much evidence is enough?

And can certainty ever be achieved?

The next essay is:

Evidence, Justification, and Certainty.