How can we deal with decision making under ignorance of probabilities when all possible negative or positive outcomes of one alternative are equal to that of the other(s)? I put forth the following example: Let's say that I can choose either to deal with a current personal security matter, which might otherwise bring about death, or to deal with a health issue that, if left untreated, might have the same consequence; and let's suppose that I have no access to the probability of mortality from any problem, nor to the probability of mortality provided that I assess either of them. As I see it, normative accounts for these instances, such as the maximin, minimax, maximax, and Laplace criteria would hold the alternatives to be equally good, as they have the same expected utility. But I am sincerely dissatisfied with the idea of making choices at random, so I want to know what you think. I also see the possibility of the decision making process being tainted by an "anything goes" type of mentality, as coming...
If I understand your question
If I understand your question correctly, it's this: in a case where the available considerations don't favor one alternative over another, how can we choose rationally what to do, where "rational" entails that anyone in the same situation (same preferences, values, information, probabilities or lack thereof...) would choose the same way?
Unless I'm missing something, you can't choose rationally in this case in that sense. The way you've set the situation up leaves no room for singling out one alternative.
One possibility is to add "do nothing" to the list of alternatives. If that's better, or likely better than each of the alternatives, do nothing. But if doing nothing is worse overall, then the obvious question is: what's wrong with picking randomly? After all, picking randomly only in cases where you need to make a choice and there's no principled way to do it isn't the same as thinking that anything goes in any circumstance.
That should be clear in general. But an artificial example may help....
- Log in to post comments