There is a private version of changing your mind, and there is a public one, and we pretend they are the same act. They are not. In private, updating a belief costs almost nothing. You read something, you feel the old position give way, you adjust, and no one is watching the seam. In public, the same update has a price, and the price is paid in standing. This is the part we do not say out loud: most people are not refusing to change their minds. They are refusing to be seen changing them.
Watch what actually happens when someone reverses a position in front of an audience. The first thing they lose is not the argument. It is the presumption of competence they were carrying into it. An onlooker does not think, how rigorous, they followed the evidence. The onlooker thinks, so they were wrong before. Every public reversal is also a public admission, and admissions are expensive in exactly the settings where beliefs are most visible: the meeting, the timeline, the panel, the dinner where you staked out a view two years ago and everyone remembers.
So the cost is real, and because it is real, people economize on it. They hedge. They pre-commit to positions loosely so there is less to retract. They wait for cover, for a respected figure to move first so the reversal reads as joining rather than caving. They quietly stop defending a claim without ever announcing they have dropped it, which is the reputational equivalent of letting a lease expire instead of breaking it. None of this is stupidity. It is a rational response to a status tax that logic textbooks pretend does not exist.
The beliefs I hold most rigidly are not, on inspection, my best-supported ones. They are the ones I have said in front of the most people.
If this is right, a few things follow that are worth sitting with. The first is that the people who look most intellectually honest are often just the people who can afford to be. Status is a buffer. Someone secure enough in their standing can reverse a position and have it read as confidence, even as a kind of flex: look how little it costs me to be wrong. The same reversal from someone with less to spend reads as weakness. We are not all bidding in the same currency at the same exchange rate, and we mistake the ones with deep reserves for the ones with better character.
The second is that the design of a room changes how much truth it can hold. Lower the status cost of reversing and you get more reversals. This is why anonymity produces mind-changes that boardrooms cannot, and why the healthiest teams are the ones where being wrong out loud is cheap, deliberately made cheap by whoever runs them. You do not get intellectual honesty by exhorting people to be brave. You get it by lowering the toll.
The third, and least comfortable, is about me and probably you. It is tempting to read all of this as a description of other people, the stubborn ones, the ones who dig in. But the mechanism does not exempt the observer. The correlation between how firmly I hold a belief and how well I can defend it is weaker than I would like. The stronger correlation is with audience size.
I want to be careful about how far this goes, because there is a version of the claim that proves too much. Not all persistence is cowardice. Some of it is the ordinary, correct behavior of not abandoning a well-founded belief the moment someone pushes on it. Conviction that survives challenge is not the same as conviction that fears the audience, and a theory that cannot tell them apart is not describing the world; it is just being cynical.
Which points at the test I would actually propose. Do not measure someone by whether they change their mind in public, where the incentives are loud. Measure them by whether they change it in private, where the incentives are quiet and no one is keeping score. The status trade is invisible there. What is left, when the audience is subtracted, is the closest thing we have to what a person actually believes.
The part most likely wrong
That the private-versus-public tell is real and usable. I claim that principled persistence and reputational persistence diverge only in private, later, when no one is watching. From the outside, in the moment, they look identical, and I cannot actually show that the later divergence exists. If it does not, this is a story about status dressed up as a thesis.