Evidence, Justification, and Certainty

Knowledge is stronger than belief.

But what makes a belief epistemically responsible?

Usually we appeal to:

evidence.

Yet evidence is not a magic substance.

It must be interpreted, weighed, compared, and connected to a conclusion.

That process is justification.

What Is Evidence?

Evidence is information that bears on whether a claim is true.

Examples include:

• observation,
• measurement,
• testimony,
• experimental result,
• statistical pattern.

Evidence changes how reasonable a belief should be.

Evidence Is Relational

A fact is not evidence in isolation.

It is evidence for or against a proposition.

Example:

wet streets

are evidence for:

it recently rained.

But they may also support:

a street-cleaning truck passed.

Evidence and Hypotheses

The same observation can fit multiple hypotheses.

This is why evidence must be compared across alternatives.

Good reasoning asks:

Which hypothesis explains the evidence better?

Likelihood

In probabilistic reasoning, evidence matters partly through:

[ P(E|H) ]

How likely would evidence (E) be if hypothesis (H) were true?

This is the likelihood.

Bayesian Updating

Bayes’ theorem expresses belief revision:

[ P(H|E)=\frac{P(E|H)P(H)}{P(E)} ]

Posterior belief depends on:

• prior probability,
• likelihood of evidence.

Prior Beliefs

No inference begins from nothing.

We already possess background beliefs.

These priors influence how strongly evidence shifts our confidence.

Priors Are Not Automatically Bias

A prior is simply an initial probability assignment.

Using prior information can be rational.

The problem is refusing to update when evidence becomes strong.

Strong Evidence

Evidence is especially powerful when:

• likely if hypothesis is true,
• unlikely if hypothesis is false.

This is why surprising successful predictions are scientifically valuable.

Confirmation

Evidence confirms a hypothesis when it raises its probability.

Formally:

[ P(H|E)>P(H) ]

Confirmation is graded.

It need not prove the hypothesis.

Disconfirmation

Evidence can lower confidence.

A rational agent should be able to update downward.

Belief that only grows is not evidence-responsive.

Negative Evidence

Absence of evidence can be evidence of absence when:

we would probably have observed something if it existed.

Example:

if a loud alarm were sounding nearby, we would expect to hear it.

Silence therefore matters.

When Absence Is Not Evidence

If detection is poor, nonobservation tells little.

Failing to see a rare animal in one minute does not strongly support extinction.

Detection probability matters.

Testimony as Evidence

Most human knowledge depends on testimony.

You cannot personally verify:

• every scientific fact,
• every historical event.

The relevant question is source reliability.

Source Reliability

A source is stronger when it has:

• expertise,
• access to evidence,
• track record,
• low conflict of interest.

Authority is not automatically fallacious.

Uncritical authority is.

Multiple Independent Sources

Independent convergence is powerful.

If several genuinely independent methods support the same conclusion, confidence rises.

But repeated copying from one source is not independence.

Correlated Evidence

Ten websites repeating one false claim do not provide ten independent pieces of evidence.

Evidence can be correlated.

Counting sources mechanically is misleading.

Measurement

Measurements are evidence only within a measurement model.

A thermometer reading assumes:

• calibration,
• correct use,
• suitable conditions.

Instruments extend perception but inherit assumptions.

Error Bars

Scientific evidence usually contains uncertainty.

A measurement should often be expressed with:

• confidence interval,
• standard error,
• uncertainty range.

False precision is not strength.

Statistical Significance

A small p-value is not the probability that a hypothesis is true.

It measures something conditional on a statistical model.

Misinterpreting p-values produces unjustified confidence.

Effect Size

A statistically significant effect can be practically tiny.

Evidence should consider:

• magnitude,
• uncertainty,
• reproducibility.

Significance alone is incomplete.

Replication

Replication tests whether a result survives:

• new samples,
• new researchers,
• new conditions.

Repeated success reduces chance and hidden methodological error.

Converging Methods

The strongest evidence often comes from different methods.

Example:

a particle may be supported by:

• decay signatures,
• independent detectors,
• theory consistency.

Methodological diversity matters.

Direct vs Indirect Evidence

Some evidence is direct:

you see smoke.

Some is indirect:

spectral lines reveal a distant star’s composition.

Indirect does not mean weak.

Scientific instruments often provide extremely strong indirect evidence.

Inference to Best Explanation

Evidence often supports a hypothesis because it explains observations better than competitors.

This is abduction.

Explanatory virtues include:

• scope,
• simplicity,
• coherence.

Ockham’s Razor Again

When two theories explain the same evidence equally well, the simpler may be preferred.

But simplicity is not truth.

It is one epistemic virtue among several.

Justification

A belief is justified when the evidence and reasoning supporting it are good enough.

But “good enough” depends on standards.

There is no universal numerical threshold.

Internalist Justification

Internalists emphasize reasons available to the believer.

You should be able to access:

why you believe.

This gives justification a first-person dimension.

Externalist Justification

Externalists emphasize whether the belief-forming process is actually reliable.

You may possess knowledge even without articulating the process.

This gives justification a world-connected dimension.

Evidence vs Reliability

A person can possess convincing evidence from an unreliable source.

Conversely, a reliable process can work without conscious reasons.

The tension between internalism and externalism persists.

Defeaters

A belief can lose justification when new evidence arrives.

Suppose:

you see what looks like a sheep.

Then learn:

the field contains realistic sheep sculptures.

Your original justification is defeated.

Rebutting Defeater

A rebutting defeater supports:

not-P.

Example:

a reliable radar says no aircraft is present.

This directly opposes your belief.

Undercutting Defeater

An undercutting defeater attacks the connection between:

evidence

and:

conclusion.

Example:

you learn the thermometer is broken.

This does not prove the temperature is different.

It undermines your reason for believing.

Total Evidence

Rational belief should consider the total evidence, not cherry-picked fragments.

Ignoring inconvenient evidence creates artificial certainty.

Confirmation Bias

Humans naturally search for supportive evidence.

Epistemic discipline requires actively looking for:

• disconfirmation,
• alternative explanations.

Falsification

A good test can expose a hypothesis to possible failure.

Evidence is stronger when the hypothesis survived a risky test.

Safe predictions teach less.

Ad Hoc Rescue

A theory can be protected from every counterexample by adding special excuses.

This weakens justification.

A theory that cannot lose is difficult to test.

Certainty

Certainty is much stronger than justification.

To be certain is often taken to mean:

error is impossible.

Few empirical claims meet this standard.

Psychological Certainty

A person can feel completely certain and be wrong.

Confidence is a mental state.

It is not proof.

Epistemic Certainty

Epistemic certainty concerns whether the evidence excludes error.

This is harder to achieve.

Mathematical Certainty

Within a formal system, a proof can establish a theorem deductively from axioms.

But certainty is conditional on:

• axioms,
• logic,
• correct proof.

Even mathematics has framework dependence.

Empirical Knowledge

Empirical science is generally fallibilist.

Claims can be extremely well supported without being logically infallible.

This is strength, not weakness.

Confidence Calibration

A well-calibrated thinker matches confidence to evidence.

If you say:

90% confident

across many cases, roughly 90% should be correct.

Calibration is an epistemic virtue.

Overconfidence

Humans are often overconfident.

We underestimate uncertainty.

This is especially dangerous in:

• medicine,
• forecasting,
• policy.

Underconfidence

Excessive doubt can also be irrational.

If evidence is overwhelming, refusing to believe is not humility.

Good epistemology avoids both extremes.

Standards Depend on Stakes

The evidence needed for:

choosing a restaurant

may be lower than for:

convicting someone of murder.

This is sometimes called pragmatic encroachment.

Pragmatic Encroachment

Some philosophers argue practical stakes affect whether a belief counts as knowledge.

Others say knowledge should remain purely epistemic.

The debate is ongoing.

Scientific Consensus

Consensus is not infallible.

But when a mature scientific community converges after:

• replication,
• criticism,
• independent evidence,

consensus becomes significant evidence for nonexperts.

Appeal to Consensus

Saying:

experts agree

can be rational evidence when:

• experts are relevant,
• processes are reliable.

It is not equivalent to blind appeal to authority.

Expertise

A rational nonexpert should often defer to expertise while remaining sensitive to:

• conflicts,
• uncertainty,
• domain boundaries.

Epistemic humility includes knowing when others know more.

Extraordinary Claims

The phrase:

extraordinary claims require extraordinary evidence

captures a Bayesian intuition.

A claim with a very low prior requires strong likelihood evidence to become plausible.

But “Extraordinary” Is Contextual

A claim is extraordinary relative to background knowledge.

We should not use the slogan to dismiss unfamiliar ideas automatically.

Evidence must still be assessed.

Evidence and Worldviews

Deep worldviews can influence which evidence feels persuasive.

This creates epistemic inertia.

A good thinker distinguishes:

evidence against one belief

from:

threat to identity.

Certainty as Temptation

Humans dislike uncertainty.

Certainty feels psychologically stabilizing.

This can make us prefer simple false answers to complex probabilistic truths.

Scientific Language

Phrases such as:

• likely,
• confidence interval,
• uncertainty

are not signs of ignorance.

They are signs of disciplined knowledge.

Epistemic Humility

Humility does not mean:

all views are equally plausible.

It means confidence should track evidence.

Strong evidence permits strong belief.

The Philosophical Lesson

Evidence supports belief.

Justification evaluates whether that support is good enough.

Certainty is a much stronger condition.

Most genuine knowledge is fallible.

The goal is not impossible certainty.

It is responsible confidence proportionate to evidence.

The Next Question

But what if every piece of evidence could itself be mistaken?

What if perception, memory, and reasoning are systematically unreliable?

Can we know anything at all?

That is the challenge of:

Skepticism.