Metasemantic Probability Nonsense

I help run a meetup called Astral Codex Ten - Tokyo. We talk about the blog Astral Codex Ten. Today we discussed https://www.astralcodexten.com/p/in-continued-defense-of-non-frequentist. We relatively quickly (by our standards - it took maybe an hour) agreed on everything but whether or not you can assign probabilities to "inconceivable" things. 

I'm very frustrated about this. We've agreed that you can, at least a little bit, estimate the probability of any particular thing, and that it's only the things you can't conceive of at all that you can't assign probabilities to - obviously not individually, but as a class. Will you, at some point in the next whatever amount of time, believe "something I thought was inconceivable has happened"?

I think the question here is about what kinds of induction are possible. We generally agreed that "the sun has risen a bunch of times in the past, so probably it will again" is reasonable. The specific thing we're in contention about is, can you say "I am a person with a model of the world, I've seen plenty of people with world models before, they run into previously inconceivable things at such and such rate" and learn something about your own likelihood of encountering previously inconceivable to you?

I think you can. I'm frustrated to find I can't get my friends to discuss this. I think the difficulty is in fully organizing your arguments in a social setting - there's a lot of people thinking different things, and if you sit and think about the proper framing of an argument for ten minutes, the conversation will leave you. I'll try writing up the best version of my argument and posting to Discord, then commit to thinking about everything I post for at least an hour before sending it.