[EM] What should an ideal single-winner method achieve?

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

[EM] What should an ideal single-winner method achieve?

C.Benham
Toby,
I want to as much as possible give every voter equal positive (from the
individual voters' perspectives, as far as we can tell from the preferences
they express on their ballots) weight in determining the result. We can
tend towards "maximising utility" by not encouraging voters to express very
weak (possibly light-minded) preferences or even arbitrary
not-really-existent preferences that the voting method is unable to
distinguish from strong
real preferences.

So I'm allergic to any sort of random-fill incentive or a rule common in
Australia that ballots that don't strictly rank all the candidates have
no effect on
the result. And I'm not a huge fan of  Later-no-Harm.  Nonetheless IRV
avoids having any random-fill incentive because it also meets
Later-no-Help and
so is invulnerable to Burial strategy and that is one of the reasons why
I rate it as an ok method.

I generally want to reduce insincere strategy incentives, and I want to
minimise the advantage strategic voters have over sincere voters and the
advantage
that well-informed strategic voters have over less-well informed
strategic voters. And I prefer that there be no zero-information (about
how other voters
will likely vote) strategy incentive but if there is it should be
intuitive and commonsensical.

There is some balance to be struck between guaranteeing the fairest
result if all the voters are sincere and on the other hand reducing
strategy incentives
so as to make it more likely that they are, so as to avoid a "garbage
in, garbage out" scenario.  For that reason some of the methods that
fail Condorcet
have some justification, because Condorcet is incompatible with
Later-no-Help (and so has some Burial incentive) and FBC.

Regarding ballot rules, I prefer if the voters are able to express a
full strict ranking of the candidates if they wish to. It is also nice
if they can express
above-bottom equal preference and  maybe also an explicit approval
cutoff or some ratings from which one can be inferred. Some limit on the
number
of candidates a voter is allowed to vote above bottom (as with plurality
and some implementations of IRV I've heard of) is unacceptable.
Approval ballots are also too restrictive.  Three-slot ratings ballots
are much better but still too restrictive. Four-slot ratings ballots are
much better again
and might be (at least most of the time) good enough.

Lastly I like to avoid  logical absurdity, such as failures of
Mono-add-plump and  Irrelevant Ballot Independence and Clone Independence.
BTW, the "strategy incentives" I referred to earlier refer not just to
voting but also to nomination of candidates.

> Obviously a voting method also needs to be simple enough to understand
> (in terms of voting and understanding how the winner is calculated)..
If you mean that the voters need to understand how the winner is
calculated, that view is not universal in the world.

> ... ,and it might be that different types of election suit different
> methods.
Yes. If the voters trust the people running the election they might put
up with a complicated voting algorithm they don't understand.  If little
is at
stake then limiting strategy incentives becomes less important and the
voters might be content with something relatively simple to save time.

Chris Benham

Toby Pereira tdp201b at yahoo.co.uk
Thu Jun 27 07:31:17 PDT 2019


> With all the discussion of different single-winner methods and the
> criteria they pass and fail, I'm interested to know what you think the
> "ideal" method should hope to achieve. For example, some people might
> want to maximise utility summed across the voters. Others might want
> to find the candidate that is closest to the "median voter". For
> others it might be more about obeying some sort of majority criterion
> (e.g. Condorcet). Etc.
> Personally, the measure that makes most sense to me is to maximise
> utility. But this doesn't automatically mean score voting (where a
> score could simply be seen as a utility rating of a candidate), at
> least in part because strategies that voters adopt might reduce its
> effectiveness. Obviously a voting method also needs to be simple
> enough to understand (in terms of voting and understanding how the
> winner is calculated), and it might be that different types of
> election suit different methods.
> Toby



---
This email has been checked for viruses by AVG.
https://www.avg.com

----
Election-Methods mailing list - see https://electorama.com/em for list info
Reply | Threaded
Open this post in threaded view
|

Re: [EM] What should an ideal single-winner method achieve?

Toby Pereira
Chris

Thank you for the response. I'll pull out a couple of quotes from your post.

>There is some balance to be struck between guaranteeing the fairest 
>result if all the voters are sincere and on the other hand reducing 
>strategy incentives
>so as to make it more likely that they are, so as to avoid a "garbage 
>in, garbage out" scenario.  For that reason some of the methods that 
>fail Condorcet
>have some justification, because Condorcet is incompatible with 
>Later-no-Help (and so has some Burial incentive) and FBC.

So what would you consider to be the fairest result if all voters are sincere? You seem to hint at Condorcet without explicitly saying it (by justifying other methods because of strategic considerations).

I think Condorcet makes intuitive sense, but actually I think becomes less reasonable the more you look into it. For example these: https://rangevoting.org/FishburnAntiC.html https://rangevoting.org/TobyCondParadox.html That's not to say I think it's completely unreasonable, but I do have doubts about it even regardless of strategic considerations. I don't automatically consider the Condorcet winner to be the "best" winner.

>Lastly I like to avoid  logical absurdity, such as failures of 
>Mono-add-plump and  Irrelevant Ballot Independence and Clone Independence.

Is participation failure not also a logical absurdity? It is effectively saying that someone's vote could count for negative. I know it's very hard to actual achieve in a voting method but it is one reason why I like score voting.





On Saturday, 29 June 2019, 13:10:56 BST, C.Benham <[hidden email]> wrote:


Toby,
I want to as much as possible give every voter equal positive (from the
individual voters' perspectives, as far as we can tell from the preferences
they express on their ballots) weight in determining the result. We can
tend towards "maximising utility" by not encouraging voters to express very
weak (possibly light-minded) preferences or even arbitrary
not-really-existent preferences that the voting method is unable to
distinguish from strong
real preferences.

So I'm allergic to any sort of random-fill incentive or a rule common in
Australia that ballots that don't strictly rank all the candidates have
no effect on
the result. And I'm not a huge fan of  Later-no-Harm.  Nonetheless IRV
avoids having any random-fill incentive because it also meets
Later-no-Help and
so is invulnerable to Burial strategy and that is one of the reasons why
I rate it as an ok method.

I generally want to reduce insincere strategy incentives, and I want to
minimise the advantage strategic voters have over sincere voters and the
advantage
that well-informed strategic voters have over less-well informed
strategic voters. And I prefer that there be no zero-information (about
how other voters
will likely vote) strategy incentive but if there is it should be
intuitive and commonsensical.

There is some balance to be struck between guaranteeing the fairest
result if all the voters are sincere and on the other hand reducing
strategy incentives
so as to make it more likely that they are, so as to avoid a "garbage
in, garbage out" scenario.  For that reason some of the methods that
fail Condorcet
have some justification, because Condorcet is incompatible with
Later-no-Help (and so has some Burial incentive) and FBC.

Regarding ballot rules, I prefer if the voters are able to express a
full strict ranking of the candidates if they wish to. It is also nice
if they can express
above-bottom equal preference and  maybe also an explicit approval
cutoff or some ratings from which one can be inferred. Some limit on the
number
of candidates a voter is allowed to vote above bottom (as with plurality
and some implementations of IRV I've heard of) is unacceptable.
Approval ballots are also too restrictive.  Three-slot ratings ballots
are much better but still too restrictive. Four-slot ratings ballots are
much better again
and might be (at least most of the time) good enough.

Lastly I like to avoid  logical absurdity, such as failures of
Mono-add-plump and  Irrelevant Ballot Independence and Clone Independence.
BTW, the "strategy incentives" I referred to earlier refer not just to
voting but also to nomination of candidates.

> Obviously a voting method also needs to be simple enough to understand
> (in terms of voting and understanding how the winner is calculated)..
If you mean that the voters need to understand how the winner is
calculated, that view is not universal in the world.

> ... ,and it might be that different types of election suit different
> methods.
Yes. If the voters trust the people running the election they might put
up with a complicated voting algorithm they don't understand.  If little
is at
stake then limiting strategy incentives becomes less important and the
voters might be content with something relatively simple to save time.

Chris Benham

Toby Pereira tdp201b at yahoo.co.uk
Thu Jun 27 07:31:17 PDT 2019


> With all the discussion of different single-winner methods and the
> criteria they pass and fail, I'm interested to know what you think the
> "ideal" method should hope to achieve. For example, some people might
> want to maximise utility summed across the voters. Others might want
> to find the candidate that is closest to the "median voter". For
> others it might be more about obeying some sort of majority criterion
> (e.g. Condorcet). Etc.
> Personally, the measure that makes most sense to me is to maximise
> utility. But this doesn't automatically mean score voting (where a
> score could simply be seen as a utility rating of a candidate), at
> least in part because strategies that voters adopt might reduce its
> effectiveness. Obviously a voting method also needs to be simple
> enough to understand (in terms of voting and understanding how the
> winner is calculated), and it might be that different types of
> election suit different methods.
> Toby



---
This email has been checked for viruses by AVG.


----
Election-Methods mailing list - see https://electorama.com/em for list info