[EM] Lotteries and Condorcet

[Kristofer Munsterhjelm]

I found this paper earlier today:
https://www.science4all.org/wp-content/uploads/RandomizedCondorcet.pdf

>From my light skimming through it, it appears to define a relation
between lotteries so that a Condorcet winner always exists, and that the
(nondeterministic) voting method that elects this CW lottery is always
individually strategy-proof and is group strategy-proof under
singlepeaked preferences.

It might be possible to generalize the majority lottery concept when
given strength of preference data so that the comparison between
lotteries take that strength of preference into account. If so, that
would provide a more natural "cardinal Condorcet" method than my STAR
generalization.

Although the method would be nondeterministic, it would be pretty simple
to define a nonzero chance of victory set and then run a conventional
Condorcet method on top. This would destroy the strategy-proofness, but
I don't know if the cardinal generalization would retain it in the first
place.

It's an interesting concept that could be useful to investigate :-)

-km