[EM] Consensus Incentive Lotteries

[Kristofer Munsterhjelm]

On 17/02/2020 01.52, Forest Simmons wrote:
> Here is a one parameter family of lotteries that encourage consensus:
> the bigger the parameter N, the greater the encouragement, while in
> every case allowing single minded minorities to have proportional
> probability for representation:
> 
> 1. The voters submit approval style ballots.
> 
> 2. The approval totals are tallied.
> 
> 3. After thorough mixing of ballots a set of N ballots are drawn at random.
> 
> 4. If all N of the ballots approve one or more candidates in common,
> then from among those the one with the greatest approval tally (from
> step one) is elected.
> 
> 5. Else the candidate approved on the first drawn ballot with the
> greatest approval tally is elected.
> 
> I need suggestions for a democratic way of deciding on an appropriate
> value of N.  In other words, how to discern the potential max consensus
> level.

You could vary N to fit an approximate supermajority requirement.
However, I think it would be better and easier (having less of a sloping
cutoff) to just make that requirement explicit:

- The voters submit approval style ballots.
- The approval totals are tallied.
- If there exists at least one candidate for which there exists a group
of x% of the whole electorate approving this candidate, then elect the
candidate among these with the greatest approval tally.
- Otherwise, choose a random ballot and pick the approved candidate with
the greatest tally.

You could use IBIFA/Relevant Ratings-type logic to reduce the problems
with a hard threshold just like those methods do.

-km