Physics Guy wrote: ↑Wed Nov 04, 2026 4:06 pm
I have a hypothesis that big changes of belief, that may get called conversion or deconversion, can be described by saying that the graph of belief versus (perception of) evidence is a sort of S-shaped curve: a zig-zag that curves in reverse in the middle. What this implies is that people are pretty rational in the lower-left and upper-right corners of the graph, where evidence and belief are either both low or both high, but that what happens in the middle is not so straightforwardly rational.
The shape of curve I have in mind is not a perfect S. In the lower left corner, it's a gentle slope upward from zero: when the evidence is overwhelmingly negative, belief is nil. If evidence becomes more positive, belief rises, just as it logically should. And in the upper right corner, likewise, the curve is a gentle slope up towards total belief when the evidence is overwhelmingly positive. If the evidence weakens a bit, belief becomes less sure, as it logically should.
In the middle, the curve becomes steeper. There's a sort of tipping-point effect, where even a small change of perceived evidence can induce a large shift in degree of belief. This might be seen as a kind of over-correction for belief perseverance (or disbelief perseverance): the upper-right and lower-left parts of the curve are actually flatter than unbiased reason would make them, but there comes a point where people suddenly realize that their beliefs are badly out of step with the way the evidence looks, and they kind of rush to catch up.
An evidence-belief curve that gets steep in this middle range would not really be rational, either, any more than the flatness of the curve at both ends. If evidence shifts across the middle range from 40 to 60, a rational being's belief would also rise from 40 to 60, but a human's belief may well rise from 20 to 80.
My idea is that this is actually rational behavior for humans, though, if we consider that the problem that humans are trying to solve rationally is not just the problem of deciding what's probably true. Most people aren't actually deciding what's true, even when we ourselves think that we're doing that. I think what we're usually doing instead is deciding what we should think about most of the time, what we should say, and how we should act. An addicted gambler who technically does know the odds, but who constantly dreams about winning, only talks about winning, and keeps on making bets, is someone whom I would say actually believes they're much more likely to win than the odds would imply, even if they can tell you the odds. So when I think about belief in this way, as a life strategy more than as a detached judgement, then it seems likely that more definite beliefs—close to zero or close to 100%—are inherently more useful as strategies, either way, than wishy-washy positions closer to 50%. So people don't want to hang around in the middle zone. They're attracted to the extremes, where nearly certain belief or disbelief offer easy decision making and other practical benefits.
But here's the weird part of my hypothesis. I'm proposing that the evidence-belief curve does more than just get steep in the middle to make up for being gentle at the ends. I'm thinking that it actually curves backwards in the middle, really like the letter S. So someone could start out seeing the evidence as overwhelmingly positive, and believing with total confidence, and then if their view of the evidence slipped down to only about 20% positive, their belief would only fall to about 80%, as they slid along the top of the S. But then what if the very same person, with the same way of deciding beliefs, instead started out seeing the evidence as overwhelmingly negative (0% positive) and had complete disbelief, before gradually changing their view of the evidence to see it as 80% positive? They might well only rise to 20% belief, sliding along the bottom of the S, because of disbelief perseverance. So there could be two people who actually think and feel in the exactly the same ways, and one believes strongly (80%) with only weak evidence (20%) while the other disbelieves strongly (20% belief) despite seeing evidence that a rational decider would consider as strong (80%). The difference between the two people is the history of their beliefs over time, and their personal history of encountering evidence, rather than any difference in the basic ways they decide to believe.
In between that upper-left 20-evidence-but-80-belief point, and the corresponding lower-right 80-evidence-but-20-belief point, is the back-curving middle part of the S. This back-curving part is flatly non-rational: somebody moving along this back-curving part of the curve is actually believing more strongly because they see the evidence weakening, or conversely doubting more just because the evidence strengthens. As the evidence changes, they're doubling down instead of reconsidering. They're living in a fantasy world in which mounting evidence against their ideas is only proving how good the conspiracy is, or whatever.
Letting adverse evidence actually strengthen belief (or unbelief) is an unstable way to think and act in the world, because beliefs are practical tools that are going to shape your decisions in life. If you're unwarrantedly confident about something (one way or the other), then you'll probably try to act on that unwarranted confidence; since the confidence is unwarranted, the consequences will likely be more adverse evidence. So if this only makes you more confident, you're going to keep getting more and more confident in your disbelief or belief. On the other hand if you're less sure about something, you won't stake much upon it, so you won't get much negative feedback, so you'll perversely become even less sure than you were, when you're on this irrational back-curving part of the S. This built-in reinforcement of confidence or downplaying will tend to fragment your beliefs, as you become more and more sure about some things and more and more non-committal about others. You become a fanatic with a few pet theories and no interest in anything else.
In practice most people can't actually stay very long on the irrational part of the curve. Their increasingly incoherent obsessions bring them to a crisis and they either jump from the S's back-curving middle to its top, or else jump down to its bottom. Someone with a low tolerance for incoherent fantasy might even jump immediately from bottom to top, or from top to bottom, without going through the back-curving middle at all, as soon as they reach the point at which the back-curving starts. This would be my theory of what ex-Mormons call the breaking of the shelf—or of a conversion experience.
This theory of the partly irrational back-curving S curve of belief as a function of evidence is of course based on decades of careful psychological and sociological research on my part ... not at all. It's nothing but a speculative analogy from something in physics (the equation of state of the van der Waals gas) which explains something that is kind of like religious conversion, namely the boiling of liquid into gas, or condensing of gas into liquid. And it has something kind of like belief perseverance, namely superheating or supercooling.
Whatever human brains do in deciding what to believe, it's obviously a lot more complicated than boiling. A pattern of behavior which is simple enough to show up in gases and liquids, however, might be a pattern that is basic enough to show up in other places as well. A liquid can remain liquid even if it is heated to well above its boiling point, as long as the heating is very gradual, the liquid is pure, and the container is clean. At that point, however, even a tiny disturbance can make the superheated liquid explosively flash into gas. This really happens with water and it can be dangerous; with other fluids, on the other hand, the sensitivity of the effect can make it the basis of a detector for subatomic particles (a "bubble chamber").
It's a complicated behavior that is in some ways perverse and surprising, but it all makes sense when you understand how the particles in the fluid are interacting. And maybe in the same way a hypothesis like this could at least serve as a metaphor for human belief perseverance, and suggest ways of understanding how and why people behave as they do. Behavior that would be badly irrational as a procedure for drawing abstract conclusions might be a viable way of selecting life strategies, taking other factors besides likelihood into account.