Evidence, Reasons, and Defeaters

By the end of this lesson, you will be able to...

• Separate a claim, a reason, evidence for that reason, and an assumption behind it.

• Identify rebutting and undercutting defeaters that should weaken a belief.

• Update a claim without either accepting the first explanation or demanding perfect certainty.

Idea in one sentence: Good reasoning asks not only “What supports this claim?” but also “What new information would make that support weaker?”

Core Insight

You are waiting at a train station. The display says:

The 18:10 train to Valencia leaves from platform 4.

You form a belief: the train leaves from platform 4.

That belief has a reason. You saw it on the station display.

Now a member of staff announces that the train has moved to platform 7. A minute later, you notice that the display is labelled “updated yesterday.”

The new information changes the situation in two different ways:

• The announcement is evidence for the opposite claim: the train leaves from platform 7.
• The old date weakens the connection between the display and today's platform. The display may no longer be a good guide.

Both pieces of information are defeaters. They do not merely add another opinion. They reduce the support for your original belief.

This is the pressure behind the lesson. A claim is not strong just because we can list one reason for it. We also need to ask whether the reason is relevant, whether its hidden assumptions still hold, and what would make us revise the claim.

The Small Situation: What Exactly Supports the Claim?

Start with the claim, not with a general feeling:

Claim: The 18:10 train to Valencia leaves from platform 4.

Now separate the parts.

These parts are related, but they are not interchangeable.

Seeing words on a screen is an observation. Treating those words as support for a claim requires an assumption about the screen. The display can be clear and genuine while still being too old for the question we are asking.

Plain meaning, scenario, precise term

Plain meaning: Evidence is information that bears on a claim. A reason is a consideration offered to support a belief.

In this scenario: The display is a reason because it is being used to support the belief that the train leaves from platform 4. The visible text is evidence only if the display is current and relevant to this train.

Technical names: A defeater is information that removes, weakens, or blocks the support a belief had. A rebutting defeater supports the opposite of the claim. An undercutting defeater weakens the link between the reason and the claim.

You do not need to decide which philosophical theory of evidence is correct before using this distinction. The practical move is to make the support and its possible failure visible.

The Initial Model

The first model is simple:

The display says platform 4.
Therefore, the train leaves from platform 4.

This model is attractive because it is fast. Station displays usually exist to tell passengers where to go. If we questioned every ordinary sign as if it were a research paper, we would miss trains while conducting an endless audit.

But the model hides a condition:

The display says platform 4
AND the display is current and reliable for this train
-> the train leaves from platform 4

The second line is not decoration. It is the bridge between the observation and the claim.

Where the Support Breaks

A new fact can attack the claim directly, or it can attack the bridge that made the reason useful.

Rebutting defeaters

A rebutting defeater gives support for the opposite claim.

Original claim: The train leaves from platform 4.
New information: The station announcement says the train leaves from platform 7.
Effect: Support now points toward not-platform-4.

The announcement may be wrong. A defeater is not automatically decisive. It is a reason to lower confidence, compare sources, or check again.

Undercutting defeaters

An undercutting defeater does not prove that the claim is false. It weakens the reason that was carrying the belief.

Reason: The display says platform 4.
New information: The display has not been updated since yesterday.
Effect: The display is weaker evidence for today's platform.

The train might still leave from platform 4. The point is that this display no longer gives us the same support it gave us before.

Scope defeaters

Sometimes a reason supports a narrower claim than the one we made.

Evidence: A sign says “Platform 4” beside one departure board.
Overclaim: Every train today leaves from platform 4.
Better claim: The board currently assigns this departure to platform 4.

The evidence did not suddenly become useless. The claim was too broad for the evidence.

A Worked Evidence Ledger

Let us update the belief step by step.

The important movement is not simply “confidence goes down.” It is more precise:

claim
-> reason
-> hidden assumption
-> defeater
-> revised claim
-> better action

The revised statement might be:

The old display first supported platform 4, but current announcements and the live board support platform 7.

That sentence preserves what was reasonable earlier without pretending the old evidence is still sufficient.

So far: A reason is not a permanent property of a belief. New information can show that the claim is false, that the reason is unreliable, or that the claim was too broad. Good inquiry keeps track of which part changed.

What a Defeater Does Not Mean

A defeater is not the same as any disagreement.

If another passenger says, “I feel platform 4 is more likely,” that is a competing opinion, not yet a strong defeater. It may prompt a check, but it does not carry the same weight as a current announcement or live board.

A defeater also does not require certainty. You can say:

The evidence for platform 4 has been weakened, and I should not act on it until I check the current board.

That is better than either extreme:

• Premature acceptance: “I saw platform 4 once, so the question is settled.”
• Total skepticism: “No display is perfect, so no evidence can guide me.”

The useful middle path is proportional revision. Ask how much the new information changes the support, how costly the mistake would be, and what quick check could resolve the conflict.

Trade-offs and Limits

Looking for defeaters improves accuracy, but it has costs.

The central trade-off is simple: more checking can protect a claim from error, but it can also delay action and consume attention.

• Checking takes time. A passenger who verifies every small detail may miss the train.
• More information can create noise. Ten vague comments are not automatically stronger than one current official update.
• A defeater can be defeated. If the staff announcement was for a different train, it no longer defeats this claim.
• Evidence can remain incomplete. Sometimes the best conclusion is “not established yet,” not true or false.
• A good reason can still lead to error. The goal is to improve the route from evidence to belief, not to promise infallibility.

You can see the boundary when the cost of checking is greater than the cost of a reversible action. In the station, staying near the information board may be enough. For a high-cost decision, the same claim may need stronger and more independent checks.

Common Confusions

Confusion: A reason is automatically evidence

Why it is tempting: if a statement explains why we believe something, it feels like support already.

Better model: A reason is offered as support. It becomes useful evidence only when it is relevant, sufficiently reliable, and connected to the claim.

Confusion: A defeater proves the original claim is false

Why it is tempting: “defeat” sounds final.

Better model: Some defeaters lower support without proving the opposite. An old display weakens the reason; it does not tell us which platform is correct.

Confusion: Any disagreement defeats a belief

Why it is tempting: disagreement feels like a direct challenge.

Better model: Compare the relevance, timing, and support behind the disagreement. Lesson 003 will examine source trust and expertise more closely.

Confusion: More evidence is always better

Why it is tempting: a longer list looks more careful.

Better model: Evidence must be relevant and independent enough to add support. Repeating the same outdated display in three group chats does not create three fresh checks.

Check Your Understanding

Check: A museum website says an exhibition is open today. At the entrance, a paper notice says the exhibition moved to another building. Is the paper notice a rebutting defeater, an undercutting defeater, or neither?

Think first, then reveal.

Answer: It is mainly a rebutting defeater for the claim that the exhibition is open at the original building. It may also support a narrower revised claim: the exhibition is open elsewhere.

Check: A friend says, “The bus will arrive in five minutes,” because a timetable from last year says so. What is the first question to ask?

Think first, then reveal.

Answer: Ask whether the old timetable is current and applies to today's route. That tests the hidden assumption connecting the timetable to the present claim.

Practice: Audit a Compatibility Claim

An AI assistant says:

Library Q is fully compatible with Python 3.13.

The only source it gives is documentation for an older major version of Library Q.

Build a short evidence ledger:

1. Write the exact claim.
2. List the current reason and the assumption that makes it relevant.
3. Name one possible undercutting defeater and one possible rebutting defeater.
4. Write a revised statement that does not overclaim.
5. Name the smallest useful next check.

A strong answer should say:

• The older documentation is relevant background, but it does not directly establish compatibility with Python 3.13.
• A missing current compatibility statement is an undercutting problem for the source's relevance.
• A failed test on Python 3.13 would be a rebutting defeater for the claim.
• A defensible revision is: “Compatibility with Python 3.13 is not established by the cited source; check current documentation and run a representative test.”

Connection to the Next Lesson

This lesson showed how new information can weaken a claim or the reason supporting it. The next lesson asks a related social question: when the reason comes from another person, what makes that source trustworthy, and where does expertise stop?

Resources

• [ARTICLE] Stanford Encyclopedia of Philosophy: Epistemology — Focus on justification, evidence, testimony, and the limits of cognitive success.
• [ARTICLE] Stanford Encyclopedia of Philosophy: The Analysis of Knowledge — Focus on why justified true belief can still be vulnerable to epistemic luck.
• [ARTICLE] Stanford Encyclopedia of Philosophy: The Epistemic Basing Relation — Focus on the difference between having a reason and actually basing a belief on it.

Key Takeaways

• Start with an exact claim, then separate the reason, the evidence, and the assumption connecting them.
• A rebutting defeater supports the opposite claim; an undercutting defeater weakens the link between a reason and the claim.
• A defeater should trigger proportionate revision, not automatic certainty or total skepticism.
• Repeating old information does not create independent evidence.
• When support is incomplete, narrow the claim, seek a better check, or choose a reversible action.