<div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><br>
But Approval has been criticized for the co-operation/defection<br>
problem. In fact, that problem is the subject of the "Approval<br>
bad-example" that I've discussed at EM.<br>
<br>
That problem can happen, but Forest Simmons proposed an excellent<br>
solution to it. Forest suggested what I call "strategic fractional<br>
ratings" (SFR). In the Approval bad-example where the A voters and the<br>
B voters greatly prefer A and B to C, and despise C, but want their<br>
own candidate to be the {A,B} candidate who wins, and if you're an A<br>
voter, and you know that the B voters would like to take advantage of<br>
your co-cooperativeness if you approved B, then you could give to B a<br>
_fractional_ rating, by probabilistically approving B. Approve B with<br>
probability that is effectively like giving B a fractional Score<br>
rating. Draw a numbered piece of paper from a bag, for example.<br>
<br>
The idea is to give to B just enough so that B will be able to beat C<br>
and win, only if the B voters are more numerous than the A voters.<br>
After all, if B is bigger, then you don't mind helping B win for you.<br>
B is then the rightful winner in {A,B}. But you don't want to help B<br>
win if B is smaller than A, and the less-numerous B voters are taking<br>
advantage of your co-cooperativeness. Hence SFR.<br></blockquote><div><br></div><div>This approach was formalized in <span style="line-height:1.35">Bouton, L., and M. Castanheira. “One Person, Many Votes: Divided Majority and Information Aggregation.” </span><i style="line-height:1.35">Econometrica</i><span style="line-height:1.35"> 80, no. 1 (2012): 43–87. Apparently since then they've done some yet-unpublished empirical work to see if people really vote that way if the conditions are right; and found that they partially do. </span></div>
<div><span style="line-height:1.35"><br></span></div><div><span style="line-height:1.35">Jameson</span></div></div>