[EM] strategy and information

[Jonathan Lundell]

On Feb 22, 2011, at 9:48 AM, James Green-Armytage wrote:
> 
> Well, I'm interested in these kinds of ideas, sort of. That is, if there are methods that give strategic incentives, but these incentives don't have a tendency to lead to harmful consequences, I'd like to talk about that. My impression is that most strategic incentives can lead to harmful consequences, but perhaps one can make distinctions between greater and lesser degrees of harm, or greater and lesser probabilities of harm --- I don't actually know, but it's a very interesting question.

To digress a little (hence the subject change), there's another dimension to looking at strategies that could use some systematic development: the way in which a strategy relates to the information a voter has about the behavior of the rest of the electorate.

The limiting cases of zero and complete information are interesting to begin with. A ranked method that gave a voter with zero information an incentive to vote other than sincerely would be perverse indeed ("elect the candidate with the most last-place rankings").

Simple plurality in the face of zero information requires a sincere vote (by "requires" I mean: requires to maximize the outcome for the voter in question). 

Approval, on the other hand, becomes problematic in the face of zero information. Suppose my sincere preference in an approval election is A>B>C. My possible (non-abstaining) votes are A or AB, but (it seems to me) the rational choice of whether to approve B requires either information about general voter behavior, or else the assignment of cardinal utilities to the candidates.

Having more than zero information moves us quickly into game-theoretic considerations of feedback, equilibria and information asymmetry. We're used to making this kind of calculation for plurality elections, and a burial strategy seems similarly straightforward.