[EM] Condorcet meeting

[Kristofer Munsterhjelm]

On 2023-08-24 14:29, Colin Champion wrote:
> One of the agenda items is "rules for reducing a large field of 
> candidates to a field of 2 to 5". This seems to me an important topic, 
> since voters cannot be expected to vote in the way ranked preference 
> methods assume if the number of candidates is large. Presumably 
> proposals have been made for addressing it; unfortunately I haven't seen 
> them.

My intuitive response is to do some type of open Approval primary. The 
incommensurability and chicken dilemma problems hit a lot less when 
there are 5 candidates to spread the error over.

Of course, the naive STAR approach is not cloneproof, which is a bit of 
a bummer.

 From one perspective, we'd *want* the "ranked general" to have a pool 
of differing shades of centrist, because the general's purpose would be 
to determine which out of the reduced pool is actually the best 
candidate. But this approach seems to be fundamentally clone vulnerable, 
i.e. it's not something you could patch up, because clones of the winner 
would all seem to be reasonable centrists in their own right. There's a 
tension between doing most of the median-finding in the primary, and 
clone independence.

So there may still be room for proportional voting type patches, even 
though they draw the general further from the idea of closely 
scrutinizing similar candidates, and even though they are somewhat ugly 
kludges.

Or perhaps there exists a better method :-)

-km