[EM] A very hard to manipulate random SPE method?

[Kristofer Munsterhjelm]

Here's an idea for an SPE method that should be pretty hard to
manipulate. It can only be done interactively.

First, the moderator chooses two candidates at random. The group votes
pairwise between the two. Like in SPE, the winner of the two is
retained. Then the moderator repeatedly chooses an unevaluated candidate
at random, pairs the current champion with the challenger, and whoever
wins pairwise goes on to the next round. The last candidate standing is
the winner

So basically it's SPE with a random candidate order, except that the
order is not fixed in advance. That it's not should make it harder still
to manipulate because it's far from obvious who a candidate should be
buried under in order for the burial to succeed.

For instance, suppose A and C are called. B>A voters know that A has
good support so they would like to bury A to create a cycle where B
wins. But they don't know if the future agenda order is set up so that
burying under C will make some loathed candidate D win.

Since it's interactive, it could also be combined with a candidate
withdrawal option. James Green-Armytage has suggested that this could
further limit strategy in Condorcet methods. In this version, after the
winner has been decided in a round, the moderator asks if any of the
candidates would like to withdraw, before continuing to the next round's
random selection.

It's not cloneproof since it uses Random Candidate instead of Random
Ballot. A better version would use Random Ballot, but it would be harder
to do so interactively - to pick a random voter and ask for that voter's
favorite among the unevaluated candidates.

And obviously there should be great security around the entropy source
so that the moderator can't control the agenda.

-km