[EM] A more efficient strategy-free ratings-based method than Hay voting

[Jobst Heitzig]

Dear Forest,

you wrote:

> If I understand your method correctly, it is a refinement of the
> following method which is based on so called "ranked preferences"
> wherein the voters have some way of expressing their own relative
> preference strengths:
>
> Three ballots i, j, and k are drawn at random. Let A, B, and C
> designate their respective top candidates.
>
> IF ballot i expresses A>C>>B AND ballot j expresses B>C>>A, THEN C is
> elected, ELSE a fair coin is tossed to decide between A and B.

Yes, I would say so.

> Would there be any incentive in this crude imitation of your method for
> voters to express false relative preference strengths?

Hm... Assuming the "ranked preferences" ballot allows me to 
simultaneously indicate A>C>>B for each pair of options A,B,C for which 
I prefer A to C and C to B and for which I prefer C to the fair 
A/B-lottery, and simultaneously indicate A>>C>B for each pair of options 
A,B,C for which I prefer A to C and C to B and for which I prefer the 
fair A/B-lottery to C, then I would think that under your method there 
would be no incentive to misrepresent any preferences or relative 
preference strengths.

Do you have in mind a particular coding scheme other than ratings? I'm 
not sure how such a scheme could look, since it would not suffice to 
indicate the relative strength of the n-1 preferences between 
neighboured pairs but would rather require to specify the relative 
strengths of all n(n-1)/2 pairwise preferences! For example, in the case
A>>B>>>C>D>>>>E
it is unclear whether the voter prefers C to the fair A/E-lottery or 
vice versa!

Yours, Jobst