> Trying to write a safety case seems like it would have bad epistemic effects on researchers in practice due to putting researchers in the frame of trying to argue why a model is safe.
I don’t really think this framing will have huge effects on the type of report that gets written. It seems like you’re worried that researchers will not write about extremely dangerous cabailities iff it’s framed as a safety case, but I doubt the nuance of calling it a risk report would incentivize the writers sufficiently to have them talk about extreme dangers more than they already do
> Trying to write a safety case seems like it would have bad epistemic effects on researchers in practice due to putting researchers in the frame of trying to argue why a model is safe.
I don’t really think this framing will have huge effects on the type of report that gets written. It seems like you’re worried that researchers will not write about extremely dangerous cabailities iff it’s framed as a safety case, but I doubt the nuance of calling it a risk report would incentivize the writers sufficiently to have them talk about extreme dangers more than they already do