[EM] Juho: Social Optimizations. The Sincere Ideal.

[Michael Ossipoff]

On 14.11.2012, at 15.21, Michael Ossipoff wrote:

> There's no "best winner". We've been over that. But, if you really
 > want a best winner, then look at the significant social optimizations
 > of Approval and Score.

There may be different elections with different needs. The society is
 free to decide what criterion to use for each need. If you want to
 elect a candidate that gets a high sum of ratings, why not use Score.
 If you want to elect the most approved candidate, use Approval.

[endquote]

Electing the most liked and trusted candidate, or the candidate
 acceptable the most people--That is a social optimization.

Or, if people vote strategically in Approval, by voting for the
 above-expectation candidates (as they will be doing no matter of which
 Approval strategy variant they use), then Approval maximize the number
 of pleasantly-surprised people. That's another social optimization of
 Approval. Because one should use Score as Approval, because Score is
 essentially Approval, I regard Score has having the same social
 optimizations that Approval has.

Juho said:

 If you want to elect a candidate that is preferred over all others,
 use Condorcet,...

[endquote]

That doesn't really describe the CW. You'd need a slightly more
 detailed wording. And the CW isn't really a social optimization, in
 the sense that Approval's social optimizations are social
 optimizations. It could be regarded as a sort of strategy benefit.

But electing the CW is a fine goal for a rank method. And, if that's
 your goal, then TUC(margins) comes close to being able to do so.

I've said that, in general, in Condorcet methods, the detailed
 information necessary to rank other than 0-info won't be available.
 But that information (though not truthful) _is_ available now, at
 top-end. And there's reason to expect that it would be, even under
 better-informed conditions. So, TUC(margins)' 0-info sincerity won't
 really operate, for that reason.

I like the CW-electing ideal too. I think that ICT will do better,
 because at least it won't make people _reverse_ preferences at
 top-end, as TUC will.

But, of course, with the optimal strategy of ranking all of the
 acceptable candidates in 1st place, ICT isn't eliciting full sincerity
 either.

That's why I say that we should reluctantly give up the
 sincere-ranking ideal, and the CW-electing ideal.

Juho continued:

 (Take also the nature of the society into account since the votes may
 not be sincere enough.)

[endquote]

Yes, that's the trouble, isn't it. But the problem is a little worse
 than Juho implied. If people vote optimally, based on what they
 believe, then, under existing conditions (and probably under likely
 future better-informed conditions too), they won't vote sincerely at
 top-end. Of course top-end is the important part.

> Maybe, then, people should reluctantly give up the
 > elusive goal of electing the CW. That's my take. Just work on reducing
 > strategy needs, eliminating the worst strategy needs.

Juho said:

I'm more optimistic. My guess is that in most societies voters are
 sincere enough.

[endquote]

Optimistic speculation isn't enough.

What is known is that, in this country, the mass-media, the tv in
 particular, have thoroughly convinced nearly everyone of two beliefs:

1. The belief that the Democrats are acceptable, in spite of their
 abysmal corruption (maybe because of #2, below)

2. The belief that only the Democrat or the Republican can win

I call that the "Democrat-disinformation".

If you believe the Democrat-disinformation, then, for you, the optimal
 strategy in traditional unimproved Condorcet (TUC)--including Beatpath
 or Ranked-Pairs--is to rank the Democrat alone in 1st place.

That just isn't controversial or debatable.

Now, you want to optimistically speculate that voters here are going to
 start voting contrary to their obvious optimal strategy, even though
 now they're following that optimal (based on their beliefs) strategy.

Sure, you can speculate, and no one can disprove an optimistic speculation.

But, what we can discuss reliably and concretely is optimal strategy,
 and the electorate's history of voting optimally (and insincerely)
 based on their media-conditioned beliefs.

Juho said:

> But, what if there are 20 or 30 candidates? Wouldn't you prefer a
 > method that doesn't make you need to rank the unacceptables?

In methods with 20 or 30 candidates many of the candidates may be
 irrelevant either in the sense that they will certainly not win, or in
 the sense that the voter doesn't care which one of the remaining
 candidates wins. In those cases truncation is quite ok. No information
 lost. It would however be good if the voters would rank all but one of
 those unacceptables that are potential winners (if the voter has such
 preferences).

[endquote]

If I ranked all of the candidates sincerely, the Democrat and the
Republican would be at the bottom of that ranking. Even if they're
winnable.

So you can't say that not ranking unwinnable candidates allows you to
vote a short ranking.

Most Condorcet versions give incentive to rank all of the candidates.
Symmetrical ICT is an exception, which I regard as an advantage (since
we're not likely to elect a CW anyway).

> if ranking unacceptables is distasteful to you (as it is to me)

You should think that you are telling that the worst candidate is even
 worse than the second worst. That's what Condorcet methods anyway
 typically do, i.e. focus on pairwise losses rather than wins. That
 could make ranking of the worst candidates a pleasant experience. :-)

[endquote]

Sure, there's a case for saying that people would enjoy indicating who
is worst. I just don't think that the Democrats and Republicans
deserve to be ranked at all. Not ranking them at all is better than
dignifying them with rank positions--even last and 2nd-to-last. Hence
my preference for Symmetrical ICT  And Approval and Score.

(But there are other reasons, as well, for preferring Approval and Score)

Mike Ossipoff




Juho