CALIBRATION · CHECKING THE INSTRUMENT
On guessing before you look — and on the finding that makes the whole first movement of this collection less impressive than it sounds.
The previous two essays asked you to take readings. Neither asked whether the instrument works.
An uncalibrated scale is not blank. It produces a figure, promptly, with no indication of error, and a person using it has no way to distinguish a good reading from a bad one except by comparing it to something.
The whole of the first movement of this collection depends on an instrument nobody has tested, and this essay is the test.
One failure declines to be measured. The other is measured into paralysis.
“I know how I am.” Sometimes true and unverifiable either way, which is the problem. A person who declines the check has a confident instrument and no error bar, and confidence is not correlated with accuracy here.
The reverse: the number becomes the reality. A reading is consulted before acting, states are experienced as data points, and the person is now managing an instrument rather than living.
A prediction made before the data is a test. The same prediction made after it is a memory, and memories of what you expected are reliably contaminated by what happened.
That is the entire method and it is not complicated. Its difficulty is emotional rather than technical: writing down a guess you might be wrong about is uncomfortable, and the discomfort is why almost nobody does it.
Confucius was asked what he would do first if given charge of a state. The expected answers were about policy, or war, or taxation.
His answer was that he would rectify the names — make the words match the things. His questioner found this disappointing, which the text records.
What makes it relevant is the reasoning he gives: if the names are not accurate, nothing said about the situation is reliable, and every action taken on the basis of it is built on a description that does not fit. That is a claim about instruments, several centuries before anybody was calibrating anything.
Interoceptive accuracy — how well a person actually detects internal bodily signals, measured objectively — is only weakly related to interoceptive sensibility, which is how good they believe they are at it.4 This is the finding that matters most here: confidence about reading yourself carries little information about whether you are reading yourself correctly.
Self-assessment generally is poor, and poorest among the least skilled, who lack what they would need in order to recognize the gap.5
And retrospective reports of one’s own states diverge systematically from concurrent ones, which is the essay-two finding arriving again from a different direction.
The limiting finding: none of this means the readings are worthless. It means they need an error bar, and that an error bar can only be got by comparison. A reading with a known bias is more useful than a reading believed to be exact.
The heroic material is not merely absent here but opposed. Its figures are magnificently unaudited; self-examination of this kind would read as a weakness in the character rather than a virtue.
The Daoist material is sceptical for a different reason — it distrusts the mind’s reading of itself in general and would regard a calibration exercise as adding another layer of the thing that is causing the trouble.
That objection is worth keeping rather than dismissing. It is the case against, below.
The strongest version is not that measurement is inaccurate. It is that measuring changes what is measured.
The case against
There is no external vantage. The prediction, the reading and the comparison are all produced by the same apparatus, which means the exercise can improve consistency without improving accuracy — you can become reliably wrong in a stable way and mistake it for calibration.
The partial answer is to anchor on something outside: an objective measure where one exists, or another person’s read. Both are imperfect and both are better than nothing.
The more honest answer is the second failure above. If checking the instrument has become a daily occupation, the instrument was never the problem, and the Daoist objection has won on the ground that matters.
Predict in writing, before looking at anything. What will the reading be? Seal it, then take the reading.
Start where an external anchor exists. Something with an objective counterpart — hours slept, a time, a count — rather than a purely internal state, because the internal ones have nothing to check against.
Look for the direction of the error rather than its size. Direction repeats and is therefore correctable; size varies and tells you little.
And set an end date. Three weeks, then stop. This is a calibration exercise, not a practice, and running it indefinitely is the capture failure.
The end date is part of the design rather than a concession.
The instrument works better than nothing and worse than you think, and the gap between those is what this essay was for.
Confidence about reading yourself is weakly related to accuracy, which means introspection cannot supply the check and something outside has to.
And it has an end date. Detection is now complete — you can read a spike, a trend, and roughly how much to trust either. Nothing in the collection so far has changed anything, and the next essay is where that starts.
Predict in writing before you look, then compare. Start where an objective counterpart exists, and look for the direction of the error rather than its size.
Three weeks, then stop. If checking the instrument has become a daily occupation, that is the finding, not the practice.
Research is cited by author and publication, with effect sizes and replication status where they matter. Where a claim rests on a single study, the essay says so.