It's sort of an interesting problem from a purely inferential perspective: for most problems, if you want to estimate the probability of X, you just go observe

It's sort of an interesting problem from a purely inferential perspective: for most problems, if you want to estimate the probability of X, you just go observe and see how often X appears. There's issues about the estimate to place with zero observations from a sample (e.g., https://en.wikipedia.org/wiki/Rule_of_succession https://en.wikipedia.org/wiki/Rule_of_succession), but basically you'd have 0 or some very small number. So why don't we just take it as 0, or ϵ? The reason is because we doubt our observational ability. So then the question becomes "is the zero because nothing is out there, or because we can't observe it?" I guess it's interesting to me because it's a salient case of where our confidence in inferential paradigms breaks down in a very obvious way. Either that, or we are uncomfortable with its implications. Or both.