[EM] Chicken Dilemma--To whom is it a problem?

[Michael Ossipoff]

Kevin--

It's true that, in your 2nd example, and in situations like it, CD can
and sometimes will falsely convict the B voters of defection,
producing a (somewhat, at least) worse result than would Beatpath, RP,
or MMPO[IA>=MPO], when they choose B.

In that example, if B voters aren't defecting, I prefer B as the
choice, because that's what RP and Beatpath choose, and I like their
all-pairwise choice.

But, as you know, sometimes different desirable properties are
mutually incompatible. I feel that the chicken dilemma can be so
destructive to cooperation and majority rule, that it's necessary to
protect against it, even at the cost of not choosing B when B is the
best choice.

After all, even if B is better, it isn't _compelling or necessary_. B
is favorite to fewest. In the matter of A vs B, A pairwise-beats B, in
addtion to having more favoriteness. Though I prefer the choice B (if
the B voters aren't defecting), it isn't quite necessary.

As you said, we can't know if the B voters are defecting. But it's
necessary to not reward them if they are, just in case they are, if we
want to protect against the potentially democracy-destructive
chicken-dilemma.

The damage that a chicken dilemma can do--to sincere voting, to
majorty-rule, to MMC, and to Condorcet Criterion, is a lot worse than
not choosing B, even though B is the best no-defection choice.

In my latest poll at Condorcet Intenet Voting Service (CIVS), the
Expanded 30-Party Poll on Party-Platforms, I like Beatpath's output
ranking better than Benham's. And, if there's no defection, I prefer
Beatpath's way of choosing too.

But it was necessary to make Benham official for the poll, just
because it's desirable to make the poll free of chicken dilemma.

CIVS offers Beatpath, RP, and Benham. When viewing the poll-results,
at the top-right of the results-page is a menu of rank-count methods.
You can view results by any of the methods in that menue (Beatpath,
RP, or Benhm).

In that menu, Beatpath is referred to as Schulze, and Benham is
referred to as Condorcet-IRV.

Greens/Green Party USA and Socialist Party USA share 1st place in the
output ranking by Benham and by Beatpath.

They differ in 2nd place, where, in Beatpath, Libertarian shares 2nd
place with Communist Partty USA.

In Benham, the Democrats came iin 12th.

In Beatpath, the Democrats came in 12th.

CIVS can be found at:

http://www.cs.cornell.edu/andru/civs.html

Once there, at the homepage, scroll down, and at the right margin
there will be a list of polls. Click on "Expanded 30-Party Poll on
Political Parties".

Michael Ossipoff



On Thu, Oct 17, 2013 at 10:36 PM, Kevin Venzke <stepjak at yahoo.fr> wrote:
> Hi Mike,
>
> First, regarding the CD criterion and SDSC: They are, it seems
> to me, almost totally incompatible. Not just incidentally but
> even in their philosophical approach to the situation. This is
> why CD makes me skittish.
>
> CD says that when the votes look like this, and A is the winner:
>
> 26 A>B
> 25 B>A
> 49 C
>
> Then in this election, B cannot be the winner:
>
> 26 A>B
> 25 B
> 49 C
>
> "Votes-only" versions of SDSC say that C can't win. So if we
> were to satisfy both CD and SDSC we would only be able to
>
> elect A (and I wouldn't see that as viable for a proposal
>
> personally).
>
>
> In the scenario where B voters truncate, SDSC essentially
>
> wants to find a majority (even hidden beneath the unviable
> A preferences) and count it if possible. This is similar in
> spirit to FBC because it means that the A voters can defeat
> C while still expressing their support for A. It doesn't
>
> harm B.
>
> But CD looks at this scenario and concludes that somebody
> deserves a beating, and the only way to do it is to punish both
> A and B voters.
>
> My concern is that I think scenario #2 is likely in the
> "near term" (given adoption of a rank method), and that it is
> likely sincere, or at least not intentionally insincere.
>
>
>
>
> ----- Mail original -----
>> De : Michael Ossipoff <email9648742 at gmail.com>
>>>  Sure. If they can enact a system that ensures they always have
>>>  incentive to vote as a mutual majority, then they don't have any
>>>  need of e.g. SDSC.
>>>
>>>  But this is a long ways off.
>>
>>>  Isn't there a sense in which it's
>>>  "more realistic," as you say above, to be concerned about whether
>>>  methods satisfy SDSC, or other criteria which could be useful to
>>>  parties that can't win in the short term but want to at least
>>>  collect their share of the votes?
>>
>> I haven't evaluated by SDSC for quite a while. I used to apply it to
>> compare some wv Condorcet methods to other methods, but I don't know
>> how Benham & Woodall do by it.
>>
>> I'm not saying that I have a monopoly on saying what's practical.
>> You're referring to a time before there is a progressive majority,
>> right?
>>
>
> Yes.
>
>>
>> Before there's a progressive majority, there isn't any good outcome
>> that we can hope for, and I feel that Plurality is the only voting
>> system that we'll have.  And are you assuming that we can enact a new
>> voting system under Republocrat rule? If we could, then, for current
>> conditions, it would be best to have one that meets FBC. I just feel
>> that if we could ever get a better voting system, it would probably be
>> _after_ electing a progressive govt, via Plurality strategy.
>>
>> I'm in the odd position of having to ask how my own criterion (SDSC)
>> applies.  ..for which conditions (current, or Green scenaio) it
>> usefully measures merit...and in what way.  I remember the definition
>> of SDSC, and that it shows some benefits of wv Condorcet. But wv
>> Condorcet would undeniably give favorite-burial need, under current
>> conditions. For Green scenario conditions then?
>>
>
> No, I meant near-term conditions.
>
> SDSC is satisfied by WV methods, yes, but it's not incompatible with
> FBC. Examples are MDDA, MAMPO, ICA, ER-Bucklin(whole). None are
> Condorcet methods though (but ICA is quite close).
>
> I suspect that you, at some point, ruled out MDDA etc. due to the
> chicken dilemma. Assuming we might agree that SDSC is useful for near-
> term elections (and maybe we can't), I guess that there isn't going to
> be one method that is both a good "near term" method and also a good
> Green scenario method.
>
> Kevin Venzke
> ----
> Election-Methods mailing list - see http://electorama.com/em for list info