[EM] Introduction (cont.)

[Richard Moore]

Roy wrote:
> but does not necessarily, degenerate to Approval. I don't think that, 
> in reality, the ability to perceive the probability of a candidate's 
> being elected gets much finer-grained than zero/close/landslide.

It doesn't have to be fine-grained. If we think X and Y are 
the front-runners, and we like X better than Y, we can fully 
approve X along with everyone we like better, and fully 
disapprove Y along with everyone we hate worse. We can scale 
the candidates we rate in between if we like, but since our 
(course-grained) assumption is that X and Y are the likely 
leaders we can probably make a similar course-grained 
assumption that the in-betweeners don't have large 
probability deltas associated with them, so our choices on 
those candidates are not going to be as important. We can 
use an above-the-mean-utility strategy on those candidates 
(using the mean of only the candidates we do think are 
front-runners -- which also works if we think there are more 
than two likely front-runners).

Richard