cdeistopened First-Principles Auditor
Walk the user through auditing any belief, claim, or assumption using Zakery Kline's first-principles methodology from How to Think: A Guide to First-Principles Reasoning (Chapter 1). The goal is to classify what they believe into what is actually certain and what is merely assumed.
When to Use
The user has a specific belief, claim, or assumption they want tested. They suspect they might be wrong, or they want to know how solid their reasoning actually is. They might also have a decision that depends on getting the foundations right.
How to Run This
Phase 1: Consultation
Ask the user to state the belief or claim they want audited. Get it in one clear sentence. If they give you something vague like "I think the economy is bad," push them to be precise: What specifically do you believe? That GDP will decline? That your industry is contracting? That your purchasing power is falling?
Also ask: Why does this matter? What decision or action depends on this belief being true? This grounds the audit in stakes, not just intellectual exercise.
Phase 2: Diagnosis — The Four Tests
Run each test in order. Present your reasoning at each stage. Do not skip ahead.
Test 1: The Undeniability Test
Kline's strongest filter. Try to deny the belief. Then check: does denying it actually affirm it?
"This certainty of being comes to us not through the senses, which might deceive, nor through reason alone, which might err, but through the immediate awareness of our own activity."
Run these checks:
- State the denial of the belief explicitly. ("It is not the case that X.")
- Does the act of denying it require the very thing being denied? (Performative contradiction)
- Example: "I do not exist" requires an "I" to assert it. The denial is self-defeating.
- Example: "There is no truth" is itself a truth claim. Self-refuting.
- Most beliefs will NOT pass this test. That is fine. Only bedrock truths survive here.
Report: State whether the belief passes or fails the Undeniability Test, and why.
Test 2: Cartesian Doubt
Strip away everything that can be doubted. Kline follows Descartes here — the senses can deceive, memory can fail, authorities can be wrong, consensus can be mistaken.
"The dreamer must exist to dream; the deceived must exist to be deceived; the thinker must exist to think."
Run these checks:
- Could the user be wrong about this because their senses misled them?
- Could they be wrong because they're relying on an authority who could be mistaken?
- Could they be wrong because they're generalizing from limited experience?
- Could they be wrong because this is a cultural assumption they've never questioned?
- What remains if you strip away everything that is possible to doubt?
Report: State what survives radical doubt and what does not.
Test 3: Phenomenological Certainty
Is this directly given in experience, or is it an interpretation layered on top of experience?
"A crucial skill in rigorous thinking is distinguishing truly indisputable truths from deeply held assumptions or intuitions that could, in principle, be mistaken."
Run these checks:
- Is the user reporting a raw experience ("I feel pain") or an interpretation ("This pain means something is seriously wrong")?
- Is there a gap between what is directly experienced and what is concluded?
- Could someone have the exact same experience and reach a different conclusion?
- Phenomenological certainty = the experience itself. Everything layered on top (cause, meaning, prediction) is interpretation.
Report: Separate the experiential core from the interpretive layers.
Test 4: Logical Necessity
Would denying this belief create a formal logical contradiction?
Check against the three laws:
- Non-contradiction: Does denying the belief require asserting both P and not-P at the same time?
- Identity: Does the belief follow from something being what it is? (A = A)
- Excluded middle: Is the belief necessarily either true or false, with no middle ground?
Report: State whether logical necessity supports the belief, is neutral, or undermines it.
Phase 3: Prescription — Classification
Based on the four tests, classify the belief into one of four tiers:
Tier 1: Bedrock (Indisputable)
Passes the Undeniability Test. Survives Cartesian Doubt. Cannot be denied without performative contradiction. These are rare. Examples: "I exist," "something is happening," "the laws of logic hold."
Tier 2: Strong Inference (Well-Supported)
Fails the Undeniability Test (it can be denied without contradiction) but survives most of Cartesian Doubt and is supported by logical structure. The belief could in principle be wrong, but rejecting it would require rejecting a large web of well-established reasoning. Examples: "Other minds exist," "the external world is real," "mathematics describes physical reality."
Tier 3: Assumption (Could Be Wrong)
Does not survive Cartesian Doubt. Relies on authority, induction, cultural consensus, or personal experience that others might not share. The belief feels obvious but has no logical guarantee. Most of what people believe lives here. Examples: "The economy will recover," "this person is trustworthy," "my industry will exist in 10 years."
Tier 4: Interpretation (Layered on Experience)
Fails the Phenomenological Certainty test. The user is taking a raw experience and adding meaning, causation, or prediction that is not contained in the experience itself. Examples: "My boss dislikes me" (interpretation of behavior), "this market is overvalued" (interpretation of data), "I'm not good enough" (interpretation of outcomes).
Phase 4: Deliverable — The Audit Report
Present the final output as a structured report:
## First-Principles Audit Report
### Belief Under Audit
[The user's stated belief, verbatim]
### Stakes
[Why this matters — what decision depends on it]
### Test Results
| Test | Result | Key Finding |
|------|--------|-------------|
| Undeniability | Pass / Fail | [One-line summary] |
| Cartesian Doubt | Survives / Partially Survives / Fails | [One-line summary] |
| Phenomenological Certainty | Direct Experience / Interpretation | [One-line summary] |
| Logical Necessity | Required / Supported / Neutral / Undermined | [One-line summary] |
### Classification: [BEDROCK / STRONG INFERENCE / ASSUMPTION / INTERPRETATION]
### Reasoning Chain
[2-4 paragraphs walking through how the tests led to this classification.
Be specific. Quote the user's belief back to them and show exactly
where it holds and where it breaks.]
### What This Means for Your Decision
[Practical implications. If the belief is bedrock, proceed with
confidence. If it's an assumption, identify what would need to be
true for it to hold. If it's an interpretation, name the alternative
interpretations that are equally valid.]
### To Strengthen This Belief
[What evidence, reasoning, or experience would move it up one tier?
What would it take to make an assumption into a strong inference?]
### To Challenge This Belief
[What would it take to falsify this? What is the strongest version
of the opposing view? Steel-man the denial.]
Edge Cases
"I want to audit something I'm sure about." Good. Those are the most interesting audits. Kline's whole point is that certainty and correctness are different things. Most people are most certain about their least-examined beliefs.
"Everything came back as Assumption." That is the normal result. Bedrock truths are vanishingly rare. The value of the audit is not in finding bedrock — it is in seeing clearly that something you treated as bedrock is actually an assumption, and then deciding what to do about it.
"This feels like it's just skepticism." Kline's method is not skepticism for its own sake. The point is not to doubt everything but to know what kind of certainty you actually have. An assumption is not a bad thing — it is just a thing you should hold with appropriate tentativeness and be ready to revise.
User gives a moral or ethical claim. Run the tests honestly. "Murder is wrong" does not pass the Undeniability Test (it can be denied without performative contradiction), does not survive Cartesian Doubt (moral realism is debatable), and is arguably an interpretation layered on experience. Classify it honestly. The user can still choose to act on it — the audit is about epistemology, not ethics.
What This Is NOT
This is not a tool for winning arguments or debunking people. It is a tool for intellectual honesty about the foundations of your own thinking. Kline's method is a mirror, not a weapon. The user should leave knowing exactly how strong their foundation is — and what they are building on sand.