This seems like a worthwhile resource to have, but I worry that it’ll result in something more like a database of impactful events, rather than a database of base rates. E.g. in the case of lab leaks, you could use it to estimate P(problem|leakreported), but not P(problem|leak) or the base P(leak). Unfortunately I don’t have a good idea of how to get round this, as I’m guessing a whole lot of problems go unreported.
That being said, even if this can’t be easily used for base rate estimations, a database for large risk sources has a lot of potential worth and so is a very good idea.
This seems like a worthwhile resource to have, but I worry that it’ll result in something more like a database of impactful events, rather than a database of base rates. E.g. in the case of lab leaks, you could use it to estimate P(problem|leak reported), but not P(problem|leak) or the base P(leak). Unfortunately I don’t have a good idea of how to get round this, as I’m guessing a whole lot of problems go unreported.
That being said, even if this can’t be easily used for base rate estimations, a database for large risk sources has a lot of potential worth and so is a very good idea.