On Tue, 2007-02-06 at 19:18 +0100, Jobst Heitzig wrote: > Dear Peter! > > Sorry to insist, but could you please show that given your new formula > it is indeed optimal to vote the true utilities? > For this you would have to differentiate the expected utility by each > expressed rating and show that all these derivatives vanish when the > ratings equal the true utilities. > I have the strong impression that, because of the normalizing constant c > which also depends on all expressed ratings, these derivatives are, > however, far from being zero, > meaning that the expected utility is in fact maximized by some set of > ratings different from the true utilities! > > Yours, Jobst
Think about the n-Substance problem again. I showed that with a square root rule, the optimal amount of each substance to buy is proportional to its utility density. Notice I said "proportional," not "equal." That proportion is c. So the normalizing constant is not new. All I did was calculate its value. You have to have a normalizing constant because the amount of credit you have available to spend is fixed. ---- election-methods mailing list - see http://electorama.com/em for list info
