[EM] Trunaction (was re: Defeat Strength)

[Kristofer Munsterhjelm]

On 9/10/22 10:48, Colin Champion wrote:
> On 09/09/2022 10:31, Kristofer Munsterhjelm wrote:
>>
>> For truncation in particular, I think equal last means that unranked 
>> candidates provides information to the method that the voter desired 
>> those candidates to be considered worse than everybody else, while "I 
>> don't know" indicates that the voting method shouldn't care at all.
>>
> I don’t agree with Kristofer's statement.

I'm not taking a position, I'm just saying that that's what the "equal 
last" interpretation implies, and contrasting that to an interpretation 
where unranked means no opinion (kitten duel example).

I'm not entirely sure which I personally think is the better 
first-principles interpretation. Practically speaking, I think Forest is 
right when he says:

> As I understand it, the custom of treating truncation/abstention the
> same as equal last rank is a practical expedient to ward off dark
> horse upsets.

That is, counting unknowns as ranked below everybody else is a way to 
prevent a completely unknown candidate from winning, similar to the 
convention of filling in unrated candidates as zero in Range. This 
suggests that if honest performance is what we care about, unranked 
should be considered "no opinion".

On the other hand, there's something intuitively appealing about the 
Plurality criterion:

"If A is ranked first on more ballots than B is ranked at all, then A's 
probability of winning should be no lower than B's",

which is about a sort of cold start problem: if the data indicates that 
some people like A, but the data is less clear about B, then that should 
be taken to be in favor of A.

If complexity is no issue, there might be a soft quorum[1] analog that 
both satisfies Plurality and doesn't presume too much about the ranking 
of unknowns. But I don't know what it would look like.

-km

[1] For range voting: https://rangevoting.org/PuzzDLaplace.html