Method notes
Method notes
Short, auditable guides to the safeguards built into the experiment register.
Each note starts with a practical problem—peeking, guessing the blind, choosing a winner after the fact—and follows it into the protocol. These are explanations of the machinery, not substitutes for the filed plans.
What preregistration prevents — and what it cannot prevent
A timestamped analysis plan closes some escape routes, but it does not turn a weak question, noisy measurement, or dishonest report into strong evidence.
Read the note →Why blinding matters
Blinding does not accuse anyone of cheating. It removes information that can steer judgement without entering awareness.
Read the note →Why a generic description feels personal
The Forer effect is not gullibility. It is a predictable interaction between flexible language, selective memory, and the promise that a text was made for you.
Read the note →Someone has to get lucky
An unlikely streak for one named person can be an ordinary event among thousands of people, cards, signs, dates, and outcomes.
Read the note →Commit–reveal randomness, explained
A cryptographic commitment cannot prove that a choice was random, but it can prove that the server did not replace that choice after seeing your guess.
Read the note →Why today’s results stay sealed until tomorrow
A live dashboard can become part of the experiment it reports. Delayed aggregates keep the crowd, the operator, and the stopping decision from chasing today’s noise.
Read the note →