Intellectual Honesty and Scientific Discipline

Every part of this project, including its purpose, moral axioms, assumptions about human nature, and proposed institutions, must remain exposed to rigorous criticism.

We owe our ideas no loyalty beyond what their reasoning and evidence justify. Uncompromising scrutiny applies to claims; it does not excuse cruelty towards people.

Standards of inquiry

• Speak plainly about weaknesses, failures, tradeoffs, and uncertainty.
• Separate value commitments, observations, inferences, and speculation.
• Make empirical claims testable wherever possible. State what would support, weaken, or overturn them; label claims that cannot currently be tested.
• Compare credible competing explanations and seek evidence that distinguishes them.
• Update confidence with evidence and explain the change.
• Avoid false precision. Use numerical probabilities where defensible; otherwise describe uncertainty and its basis.
• Publish methods, sources, limitations, and negative results wherever privacy and participant protection permit.
• Evaluate actual consequences, including hidden costs, unequal burdens, and opportunities for abuse.
• Apply scrutiny to founders, maintainers, supporters, and accepted principles.
• Treat admitting error and revising a position as responsible contributions.

Bayesian discipline

For consequential claims, record:

1. What we currently believe and why: our prior assumptions and confidence.
2. The strongest competing explanations.
3. How likely the evidence would be under each explanation.
4. The reliability and independence of that evidence.
5. How confidence should change.
6. What further observation would best distinguish the alternatives.

Where meaningful probabilities can be assigned, posterior odds equal prior odds multiplied by the likelihood ratio of the evidence under the competing hypotheses. Evidence that is equally expected under both does not favour either one.

Priors, models, and evidence selection can encode preferences. They must remain visible and contestable. Do not count repeated reports of the same underlying evidence as independent confirmations. Examine whether plausible alternative priors or models change the conclusion.

Emotions, evidence, and judgment

An emotion is a real experience; the interpretation attached to it may be wrong. Feeling threatened establishes that someone feels threatened. It does not, by itself, establish that another person poses a threat.

Fear does not prove danger. Anger does not prove injustice. Disgust does not prove wrongdoing. Confidence does not prove truth. Collective outrage does not turn an allegation into evidence.

Emotions can reveal suffering, needs, values, and possible problems. They should prompt investigation. Claims about the world still require evidence and reasoning. Personal testimony is evidence to evaluate fairly; emotional expression alone neither proves nor discredits its factual content.

Apply this standard equally to ourselves, allies, and opponents. Sincerity and intensity do not make a belief accurate. Emotional detachment does not guarantee sound judgment, and dismissing someone as emotional does not refute their evidence.

Acknowledge the experience. Examine the interpretation. Test the claim.

Limits and responsibility

Science can test claims about behaviour and institutional consequences. It cannot, by itself, establish which moral values we ought to adopt. Logical consistency does not establish moral truth; a successful experiment does not establish universal applicability.

Decisions must account for uncertainty, consequences, reversibility, and the rights of those affected. Where evidence is weak, favour bounded experiments with explicit stopping conditions and appropriate consent.

The framework must be capable of discovering that its own foundations are mistaken and correcting them.