I think Conor means specifically that instead of saying “P” you should say (and think) “P, with probability p”. In these cases, you’re in no danger of taking p=0, or indeed of taking p<1/2.
(Of course I agree—and I’m sure Conor does too—that in general, when stating probabilities, you want to avoid probabilities of 0 and 1. Though in some extreme cases I think the actual downside of treating the probabilities as 0 or 1 is negligible, and doing so is a useful optimization.)
I think Conor means specifically that instead of saying “P” you should say (and think) “P, with probability p”. In these cases, you’re in no danger of taking p=0, or indeed of taking p<1/2.
(Of course I agree—and I’m sure Conor does too—that in general, when stating probabilities, you want to avoid probabilities of 0 and 1. Though in some extreme cases I think the actual downside of treating the probabilities as 0 or 1 is negligible, and doing so is a useful optimization.)
Endorsed. =)