Consider the principle that if an action benefits some and harms none, then it's permissible. Now imagine a lottery run by uniformly choosing a random number between 0 and 1, with each number equally likely. There are infinitely many tickets, each bearing a different number between 0 and 1. Each ticket has been sold to a different person (there are lots of people in this story!). At night, I steal all the tickets. I then rearrange them in the following way. I get all the tickets numbered between 0 and 0.990, and my best friend gets all the tickets numbered between 0.990 and 0.999. I then redistribute the remaining tickets to all the people who bought tickets. So by morning, my friend and I have all the tickets numbered between 0 and 0.999, but everybody who had a ticket still has a ticket, and the ticket she has is just a good as the one she had before. I have made it pretty sure that I would win, but I haven't lowered anybody else's chances at winning.
Bracketing contingent considerations of public peace and of positive law, it seems that:
- I have harmed none—no one's chance of winning has gone down.
- I have benefited some myself and my friend—our chances of winning have gone up.
- I have done wrong.
- Thus, an action that benefits some and harms none can still be wrong.
One might object. Suppose ticket number 0.458 wins. Previously it was assigned to Mr Smith. Now it's mine. Haven't I harmed Mr Smith? Maybe but maybe not. Let me fill out the case by saying that there is no fact of the matter as to who would have won had I not shuffled the tickets. In the story, we live in a very chaotic universe, and any activity—be it stretching one's arms in the morning or shuffling tickets—affects the random choice of winning number. There is no fact about what that random choice would have been had things gone differently. Thus just because ticket 0.458 wins and it was Mr Smith's before my night-time activity one cannot say that I have harmed Mr Smith. (Molinists won't like this. But surely whether I have harmed anybody shouldn't depend on the truth values of Molinist conditionals.)