Chapter 6: Questions, Claims, and Evidence

In the last chapter, you saw how your mind actually behaves.

You saw that it predicts, rather than just receives; that it runs on grooves and shortcuts; that it protects certain beliefs as if they were parts of your body; and that most of what you "know" comes through other people as much as through direct experience. This is the living system your epistemology has to work with.

Now we start turning that into a toolkit.

Before we can build anything more elaborate—Null Hypothesis, Burden of Proof, falsifiability, confidence as a gradient—we need to get clear on three basic building blocks:

• What is a question?

• What is a claim?

• What counts as evidence?

These sound simple. They are not. Most real‑world epistemic messes begin because these three have been blurred, swapped, or never named at all.

Questions: what you are really asking

A question is not just a sentence with a question mark. It is a request for change in your map. It says: "There is a gap here, or a tension, or a doubt. I want something that could fill it or resolve it." The quality of what you get back depends heavily on the quality of what you ask.

There are at least three kinds of questions that show up in epistemological practice:

1. Clarifying questions.These ask, "What exactly is being talked about?" They clean up vagueness.

   • "When you say 'AI', do you mean current large language models or hypothetical future minds?"

   • "When you say 'safe', are you talking about short‑term side‑effects or long‑term risks?"

2. Substantive questions.These ask about the world: "What is the case?" or "How does this work?"

   • "Does this treatment reduce mortality in this group?"

   • "Did this event actually happen the way it is being described?"

3. Normative questions.These ask about value: "What should we do?" or "What is good, fair, just?"

   • "Should we deploy this system at scale?"

   • "Is it acceptable to use this data in this way?"

Problems begin when you ask one kind of question and treat the answer as if it were answering another.

For example, you ask a normative question ("Should we approve this drug?") and accept a purely substantive answer ("It reduced blood pressure in one small study") as if that settled the should. Or you ask a substantive question ("Did this happen?") and receive only values‑talk in return ("Our critics are bad people"), which never touches the actual point.

Epistemological skepticism starts with noticing what kind of question is actually on the table—and insisting, kindly but firmly, that answers speak to that.

A quick exercise: think of a recent argument you were in. What were you each really asking? Were you aligned on the type of question?

Claims: what is actually being asserted

If a question is a request for change in your map, a claim is an attempt to produce that change.

A claim is a statement that can, in principle, be true or false, well‑supported or weakly supported. It says: "Here is how I think the world is" or "Here is what I think will happen." You cannot evaluate a claim you cannot clearly state.

In practice, many "claims" arrive entangled with emotion, metaphor, or vagueness. For example:

• "This AI system is dangerous."

• "The media is lying."

• "People today are more fragile than they used to be."

Each of these hides multiple possible precise claims inside it. "This AI system is dangerous" could mean:

• It sometimes produces incorrect answers.

• It increases the risk of certain kinds of harm (misinformation, scams, cyber‑attacks).

• It could, in some hypothetical future, escape human control and cause catastrophic damage.

These are very different claims. They require different evidence, different burdens of proof, and different responses.

A core move in epistemological skepticism is to extract a testable claim from a vague one. You might ask:

• "Dangerous in what sense? To whom? On what timescale?"

• "If your view is right, what would we expect to see? If it's wrong, what would we expect instead?"

Once a claim is precise enough, you can start asking the next questions: "What evidence supports this?" "What evidence counts against it?" "How strong is the overall case?"

The more important the decision riding on a claim, the more work it is worth doing to make the claim precise.

Evidence: what counts, and when

If a claim is an attempt to change your map, evidence is whatever bears on whether that change is warranted.

In plain language: evidence is anything that should move your confidence up or down, if you are being honest.

This sounds generous. It is not. It excludes more than it includes.

Evidence is not:

• Pure assertion ("Believe me, I know").

• Sheer repetition ("Everyone is saying…").

• Tone ("They sounded so confident/angry/eloquent").

• Identity alone ("People like us think…").

Those can be signals about where evidence might be. They are not evidence by themselves.

Instead, evidence has at least three features:

1. It is about the world, not just about your feelings.A measurement, an observation, a record, a behaviour, an outcome.

2. It is connected to the claim by a plausible story.If the claim is "this medicine reduces headaches," then "people who took it reported fewer headaches in a trial" is relevant; "the CEO seems trustworthy" is not.

3. It is, in principle, checkable.Someone else could, at least in theory, look where you looked or trace how you got from data to interpretation.

This does not mean that only lab studies count. Testimony can be evidence. Lived experience can be evidence. Historical records can be evidence. But in every case, the question is: "Why should this observation move my confidence? How strong a shove does it deserve?"

Later, in Chapter 9, we will talk about an evidence hierarchy: not all evidence is equal. For now, the important move is simply separating "this is a strong feeling or assertion" from "this is something that actually bears on the claim."

Fact vs interpretation, data vs story

One way evidence gets muddled is that facts and interpretations are blended into a single narrative.

Fact: "In a trial of 1,000 people, group A had 50 heart attacks, group B had 40."Interpretation: "Therefore, treatment B is clearly safe and effective, and critics are fear‑mongering."

Fact: "This model produced incorrect answers in 12% of test cases."Interpretation: "Therefore, AI is fundamentally unreliable and should be banned."

The factual parts can often be checked: numbers, events, quotations, dates. The interpretations go beyond them: what caused what, what it means, what we should do.

In conversation and media, these are constantly fused.

Epistemological skepticism trains you to separate them:

• "What are the raw facts here, as best we can tell?"

• "What interpretations are being layered on top?"

• "Are there alternative interpretations that fit the same data?"

Similarly, data and story are entangled.

Data: specific observations (measurements, reports, recordings).Story: how someone strings those observations together into a narrative of cause, meaning, or moral.

You need stories; no one lives on raw data. But stories can be compelling even when they are badly aligned with the data. Part of the work ahead will be learning to enjoy good stories while still asking, quietly, "Is this the only story that fits these facts?"

Putting the pieces together

By this point, you can already see how these building blocks interact:

• A question frames what you are looking for.

• A claim proposes an answer: "The world is like this."

• Evidence bears on whether that proposal is any good.

Misfires happen when:

• The question is vague or unshared.

• The claim is too fuzzy to test.

• Evidence is replaced by assertion, identity, or story alone.

In the next chapters, we will layer the tools of epistemological skepticism onto this structure.

• The Null Hypothesis will give you a starting stance for claims: "not yet persuaded."

• The Burden of Proof will help you decide who needs to provide evidence, and how much.

• Falsifiability will sharpen "what would count against this?"

• Confidence and proportional scrutiny will help you connect how much evidence you need to how much is at stake.

Those tools will not make any sense if questions, claims, and evidence are still blurred together.

A small practice: three passes on a headline

Here is a simple exercise you can start using today.

Next time you see a striking headline or social‑media claim, take 30 seconds and do three passes:

1. Question pass.What question is this headline implicitly answering? ("Is this policy working?", "Is this person trustworthy?", "Is this technology safe?")Is that the question you actually care about?

2. Claim pass.What is the claim, stated in one clear sentence?Could you, in principle, imagine evidence for and against it?

3. Evidence pass.What evidence is actually being offered? Numbers? Anecdotes? Expert quotes? Vibes?How much should this move your confidence, given the stakes?

You do not have to reach a conclusion each time. The point is to start seeing the structure.

Once you can see it on the surface—in headlines, conversations, arguments—the deeper tools of epistemological skepticism will have something solid to work with.

Next: Chapter 7 – The Null Hypothesis and the Burden of Proof