The greater the entropy of a lottery, the harder to predict who will win the election. The greater the cooperation among the voters, the easier to predict the winner. If there is near total consensus, then the winner can be predicted almost surely, etc.
This suggests trying to minimize entropy (subject to some kind of natural constraints) as a way of finding lotteries that incorporate cooperation. Here's one idea along these lines: Let's say a lottery L is feasible iff each ballot supports (with its full quota of probability) a candidate that is rated at or above the ballot's expected rating of the winner according to L. The winning lottery is one that has minimal entropy among feasible lotteries. Could this lead somewhere? ---- Election-Methods mailing list - see http://electorama.com/em for list info