[EM] [Fwd: Re: Strong Favourite Betrayal Criterion at last!]

[Forest Simmons]

 On Thu, 3 Apr 2003, Chris Benham wrote in part:
>
> I see..with your method a voter would have no reason not to approve
> Favourite vs. Lesser  Evil.
> I think that it should include a default "bullet-vote" feature : that if
> a voter just ranks one candidate 1, then that voter is automatically
> approving all the pairs that include that candidate.

I've thought some more about this. Perhaps the default should be the zero
information optimal strategy.  Then any information from the polls could
be used to tweak the default, rather than starting from scratch.

What would zero information strategy be?

For each pair find the average of the utilities of its two members. If
this average is better than the average of the utilities for the ballot
(estimated from the ballot rankings or ratings), then approve the pair.

In the case of rankings this would mean approve all pairs such that the
sum of ranks is better (i.e. smaller) then the sum of the highest and
lowest ranks on the ballot.

Forest