In this kind of situation, a cost-benefit analysis is difficult or impossible since the benefits can't really be quantified.
In practice, people often decide if something is too risky by observing if authority figures or other people in general consider the risk acceptable or unacceptable. But it's not always easy to tell that. For example, the recommendations regarding physical distancing during the pandemic are not clear to me, and the situation is too new for there to be any established norms, at least for what I have noticed.
I haven't been able to find any resources on the subject which consider something other these two approaches. Besides, I think it should be possible to derive an acceptable probability for taking a risk with zero quantifiable benefit, even if the number is not exact. If experts can agree on recommendations for what is acceptable risk, then that can't be entirely arbitrary. Would it be something like one in a billion, or one in a million, or what?