Epistemic Posture
Every claim comes with an implicit confidence level. The reader is always asking: how much should I update my beliefs based on this? Your epistemic posture — how you signal certainty, uncertainty, and the basis for your claims — determines whether they can answer that question.
Bad epistemic posture makes everything sound equally certain. Good epistemic posture lets the reader calibrate.
Carl Sagan’s “baloney detection kit” models this explicitly. He separates what we know from what we suspect from what would change our minds. The reader isn’t just receiving claims; they’re receiving a map of the claim-space with confidence contours.
Feynman goes further: he treats epistemic honesty as the foundation of science itself. In “Cargo Cult Science,” the core ethic is that you must not fool yourself — and you’re the easiest person to fool. This shows up as a style: he points to what would falsify the claim, shows the moving parts, refuses to hide unknowns.
The opposite is epistemic cosplay: using the language of certainty or uncertainty without the underlying calibration. Hedging everything to avoid commitment. Or asserting everything with false confidence to seem authoritative.
Posture operates at multiple levels:
Claim-level. Is this specific statement well-established, probable, speculative, or contested? “We know that…” vs. “Evidence suggests…” vs. “One possibility is…” vs. “Experts disagree about…”
Source-level. Where does this come from? Direct observation, inference, expert consensus, a single study, personal experience, speculation? Naming sources isn’t just citation — it’s calibration.
Model-level. Is this a description of reality or a useful simplification? Kahneman’s two systems aren’t neuroscience; they’re a framework. Making that explicit changes how the reader holds the idea.
Meta-level. What’s your overall relationship to uncertainty? Are you someone who overclaims, underclaims, or calibrates well? Readers develop priors about authors.
Steelmanning is epistemic posture in action.
Scott Alexander often states opposing views better than their proponents. This isn’t just fairness; it’s signaling. It says: I’ve actually engaged with the counterargument, and I still believe my position. The reader knows that surviving objections have genuinely been considered.
The opposite — strawmanning or ignoring objections — signals either ignorance or bad faith. The reader can’t tell if the author’s confidence is earned or naive.
Posture failure modes:
Uniform confidence. Everything stated with the same certainty. The reader can’t distinguish established facts from speculation. Academic writing often does this — hedged phrases become verbal tics rather than actual uncertainty signals.
Defensive hedging. “Perhaps,” “arguably,” “it could be said” on every sentence. This isn’t epistemic humility; it’s avoiding commitment. The reader learns nothing about your actual beliefs.
False expertise. Speaking authoritatively on topics outside your competence. Sometimes revealed by getting the easy parts wrong; sometimes only caught by domain experts.
Overconfident uncertainty. “No one knows” or “experts disagree” when actually there’s reasonable consensus. Feigned epistemic humility can be as misleading as false confidence.
The goal isn’t maximum hedging. It’s accurate signaling.
Some things are known. State them cleanly. Some things are uncertain. Say how uncertain and why. Some things are contested. Represent the disagreement honestly.
The reader should finish knowing not just what you believe, but how confident they should be in your confidence.
Related: [[explanatory-writing]], [[risk-vs-uncertainty]], [[models]], [[useful-fictions]], [[satisficing]]