Nine of the ten traits in the closeness module carry a directional prediction, written into the item bank before any respondent existed.
The tenth says this:
NO DIRECTIONAL PREDICTION. Whether shame is sharper before strangers or before people whose judgement you cannot escape is genuinely open, and either answer is informative.
That is Self-Consciousness, and I want to defend leaving it blank.
Both stories are good
Here is the case for it being sharper with strangers: you do not know how you are being read, you have no history to fall back on, every signal is ambiguous, and embarrassment is the feeling of being evaluated by someone whose evaluation you cannot predict.
And here is the case for the opposite: a stranger's opinion evaporates. You will never see them again. The people whose judgement actually costs you something are the six people who will still be there next year, and whose disappointment you would carry.
I find both of those completely convincing, which is exactly the condition under which you should not write down a prediction.
Why not just guess
Because a guess that turns out right looks identical to knowledge afterwards, and I would have no way to tell myself apart from someone who understood the mechanism.
The predictions in this instrument are load-bearing. They are what make the structural claim able to fail. Diluting them with a coin-flip I could later describe as insight would cost the other nine their meaning.
There is also a practical failure mode I have already walked into. A recruiting post I wrote described Self-Consciousness as one of two facets predicted to run backwards, and reported a confirmed reversal on it. The item bank makes no such prediction — the post was claiming a successful prediction that had never been made, in the piece aimed at the audience most likely to open the bank and check. It is fixed, and the lesson is that "we predicted nothing here" is a state that needs defending, because prose drifts toward tidier stories on its own.
And then somebody answered it
A reader reported that their entire gradient loaded on Self-Consciousness — and that theirs ran higher with acquaintances than with strangers. Not a straight line in either direction: a hump in the middle.
That is one person, and one person settles nothing. But it is the first real data on the one question I declined to answer in advance, and it arrived in a form neither of my two stories predicted. Which is roughly the best thing an open question can do.