Well small positive probabilities need not be finite if we have a non-archimedean utility framework.
Infinidesimal times an inifinite number might yield a finite number that would be on equal footing with familiar expected values that would trade sensibly.
And it might help that the infinidesimals might compare mostly against each other. You compare the danger of driving against the dangers of being in a kitchen. If you find that driving is twice as dangerous it means you need to spend half the time to drive to accomplish something rather than do it in a kitchen rather than categorically always doing things in a kitchen.
I guess the relevance of waste might be important. If you could choose 0 chance of death you would take that. But given that you are unable to choose that you choose among the death optimums. Sometimes further research is not possible.
Well small positive probabilities need not be finite if we have a non-archimedean utility framework.
Infinidesimal times an inifinite number might yield a finite number that would be on equal footing with familiar expected values that would trade sensibly.
And it might help that the infinidesimals might compare mostly against each other. You compare the danger of driving against the dangers of being in a kitchen. If you find that driving is twice as dangerous it means you need to spend half the time to drive to accomplish something rather than do it in a kitchen rather than categorically always doing things in a kitchen.
I guess the relevance of waste might be important. If you could choose 0 chance of death you would take that. But given that you are unable to choose that you choose among the death optimums. Sometimes further research is not possible.