Constantine Sandis and Nassim N Taleb in The Philosophers' Magazine Online:
There are facts that, through no fault of our own, we cannot help but be ignorant of. For example one cannot be expected to know where the next plane crash will take place, but it may be more rational to board an airline with a good record. Let us call this lack of knowledge justified ignorance. Actions affected by such ignorance are risks performed under justified uncertainty. Philosophers will quibble over whether we can ever know anything for certain, but we can all agree that many actions performed by mere mortals involve such risks.
The precise degree of risk and justification is dependent upon the availability of relevant information, and so will vary wildly from one case to another. This need not concern us much here for the rule we shall propose is intended to hold across all cases. Indeed, we maintain that the true measure of risk should not be calculated in terms of the pure probability of outcomes but multiplied by the significance of the outcomes in question. Risking losing one dollar against the ridiculously low chances of winning the lottery is far more prudent than taking a nuclear action that has a 99% likelihood of not ending the world. This is particularly true when it comes to sequences of risky decisions where a small probability of extinction, taken repeatedly, ends up raising the odds to close to certainty.
In moral philosophy there is a famous debate about the relation of duty to ignorance. Some argue that our obligations are tied to how things actually are (or will be), others to how we happen to think they are, and others still to how we can rationally expect them to be, given the information at hand. These views are all united by the thought that there is one right answer to this question of duty. An alternative school of thought maintains that there are several different obligations: the objective “ought”, the subjective “ought”, the “ought” of rational expectation, and so on. We shall not concern ourselves with these questions in this essay, important though they may be. Instead, we shall introduce a normative constraint which cuts across them in the sense that it holds true no matter which of the above views is the correct one to take.
We propose that actions performed under justified uncertainty should be subject to the Silver Rule (SR):
Do not expose others to a harm the near equivalent of which you are not exposing yourself to.