[EM] STAR

[C.Benham]

Toby Pereira wrote:

> I'm not a fan of STAR, but I am still interested in seeing how it 
> stands up to scrutiny given that it has a following. (Actually I'm not 
> aware of how STAR fails monotonicity. I was under the impression that 
> it passed.)
>
Toby,

To give you a bit of a preview before I get around to cooking up all the 
examples, nothing with such obvious Push-over incentive can meet 
mono-raise (aka "monotonicty")

Suppose  X beats Y in the final.   Now suppose on some ballots with Y 
above X, we raise X so it is now above Y.  That could reduce Y's score 
enough for it to be replaced in the final
by Z, a candidate that pairwise beats X.

Voters who are mainly concerned to have their favourite X win and are 
fairly certain that X will reach the final will have a strong incentive 
to give X max points (5) and then also
give a 4 (or even a 5) to all those candidates that they think X can 
beat pairwise.

If enough voters use that strategy and it fails, both the finalists 
could be candidates with little sincere support.

Chris Benham

On 15/08/2023 9:38 pm, Toby Pereira wrote:
>
>
> On Tuesday, 15 August 2023 at 06:28:51 BST, C.Benham 
> <cbenham at adam.com.au> wrote:
>>
>> >>As for what the boxes are about, they are about ensuring that 
>> voting methods have sensible behaviour in certain situations, so I 
>> wouldn't expect them to >>necessarily negatively correlate with a 
>> "good" result.
>>
> >Then you are not using the right boxes.
>
> I think you probably misread my sentence. I would *not* expect them to 
> *negatively* correlate with a good result. So I probably would expect 
> passing criteria to correlate positively with a good result. Apologies 
> for the convoluted sentence.
>>
>> >>It's what I just replied to Kristofer - the fact that there's no 
>> way that you can consistently define society's preference in a way 
>> that you can determine whether >>society prefers A or B by looking at 
>> the pairwise comparison.
>
> >Then what do you "look at" ??
>
> There's different things you can look at. But what I said is true, not 
> just my opinion, so it's a question for everyone. But you can still 
> look at pairwise comparisons and indeed use a Condorcet method. But 
> I'd want someone to come to this by evaluating various methods and 
> deciding that this is the best compromise overall, rather than 
> deciding from the outset that this is the One True Way. I'm not 
> anti-Condorcet, if this is how it's come across.
>
>> >>I would counter this by saying that it's based on this logical 
>> contradiction.
>
> >I'm still not seeing this "logical contradiction".
>
> It's just the premise that if A pairwise beats B then society must 
> prefer A to B. But it makes no sense to say that society prefers A to 
> B, B to C and C to A. I don't think this is controversial. It's a 
> centuries-old realisation.
>
>> >>I would also counter it by saying that someone could equally say (perhaps have a greater 
>> claim) that a method cannot be democratic if it fails participation
>
> >I don't see that either. So which method that meets Participation do 
> you like?
>
> I do actually like both score and approval. I like their simplicity 
> and the fact that a complete numerical results list can be published 
> that's simple and easy to understand. And I like the fact that there's 
> no hidden weird paradoxes in either method. Any criterion they fail is 
> pretty obvious.
>
> My absolute favourite method isn't suitable for many elections because 
> it has to be done online, but would be suitable for votes in online 
> communities etc. It's approval voting but where the current scores are 
> visible and where a voter can change their vote as much as they like 
> until the deadline. But because co-ordinated factions might try to 
> mess with this by voting or changing their vote at the last minute, I 
> think a non-deterministic end point might be necessary. So you might 
> have a week of voting plus and end section that has a half life of one 
> hour. There may be a sense in which this fails participation though 
> because with perfect game theoretical voting it should become 
> Condorcet. However, this is likely to be from other voters' reactions 
> to your presence in the voting procedure rather than you voting 
> against your own interests. This is a fairly nebulous failure and one 
> I could easily tolerate for a clean results table and a transparent 
> and simple method. Also it might in practice not always elect the 
> Condorcet winner if the Condorcet winner isn't particularly liked. E.g.
>
> 49 voters: A>>C>B
> 49 voters: B>>C>A
> 2 voters: C>A>B
>
> C (the Condorcet winner) is unlikely to get off the ground in the 
> first place in this case.
>
>>
>> >>Two of the main ones I tend to look out for are monotonicity and 
>> independence of clones because they are obviously things we would 
>> want and they don't >>seem to be too restrictive in terms of methods 
>> they allow
>
> >STAR fails both of them, "badly".
>
> >STAR is obviously garbage and a strategy farce,  as I'll explain 
> later on EM.
>
> I look forward to seeing it. As I say, I'm not a fan of STAR, but I am 
> still interested in seeing how it stands up to scrutiny given that it 
> has a following. (Actually I'm not aware of how STAR fails 
> monotonicity. I was under the impression that it passed.)
>
> Toby
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20230817/fb8a4ea5/attachment-0001.htm>