[EM] Let's clear up some confusion

[Juho Laatu]

On 4.10.2012, at 7.49, Michael Ossipoff wrote:

> You said:
> 
> , maybe the question if you recommend the voters to rank sincerely or
> if you recommend them to sometimes use the top ties (although the
> candidates are not equally good).
> 
> [endquote]
> 
> Good question. In a public election, I'd emphasize that their best
> strategy in a u/a election is to equal top rank all of the
> acceptables, and to not rank any accceptables.

Ok, this is maybe where our thouhts differ. My target level with ranked methods is to be able to tell the voters that there indeed are some strategic possibilities, at least theoretical ones, but for them it is safe enough and their best general strategy to rank the candidates sincerely. I don't like the idea of recommending voters not to give their opinion about the candidates. Not giving one's sincere opinions may infroduce more problems than it solves. And it certainly makes the outcome of the election worse in the case that there was after all no need to vote strategically.

Maybe you don't worry about recommending voters to turn strategic in Condorcet elections since you dont like (or trust) Condorcet methods very much anyway. I think that a considerable part of the benefits of ranked methods would be lost if people would have to resort to strategic voting, and not indicate sincerely which candidates are good.

Maybe your answer is that Condorcet methods are so vulnerable to strategies that the voters must either vote strategically or fall prey to the strategists. I guess we'll just have to wait and see. There are many TTR and IRV elections where people generally vote sincerely, but one never knows. Of course this depends to some extent also on the society in question.

> in the unimproved Condorcet versions, you'd never really
> know what to do. You'd have to try to judge whether some acceptables
> are much more winnable than others, so that refusing to top-rank Y (an
> acceptable) is justified because Y isnt likely to win anyway, and
> top-ranking hir could make X (another acceptable) lose.
> 
> That's a problem that you'd never have to worry about in Symmetrical
> ICT. It's the old lesser-evil dilemma of FBC-failing methods.

As I said, to me the most interesting strategic thinking related question in Condorcet is if one can honestly recommend sincere voting as the best general strategy to the voters. In Symmetrical ICT, with your advice to the voters, voters would have to worry whether to tell that their favourite is better than the compromise or if they should vote for a tie (and not support their favourite over the compromise).

> You might say that Symmetrical ICT's u/a strategy sounds a lot like
> Approval. Yes, and the fact that the best that you can get, in u/a
> strategy, in rank methods, is simply Approval strategy  is another
> good reason to propose Approval (or maybe Score) instead of any rank
> method.

Maybe you promote Symmetrical ICT because that helps you promoting Approval over the Condorcet methods, i.e. that there is a need to flatten your preferences to the level of Approval anyway. :-)

More seriously, my point is still that the best promise of Condorcet methods is that people might be able to give full sincere rankings in competitive elections. And that (e.g.) FBC properties of some traditional Condorcet methods may be good enough to allow voters to vote sincerely. Maybe you think the same way, except that for you the line of defence is at the level of Approval.

> You asked me how I'd instruct a voter. But you know that I don't
> propose any rank method for official public elecions.

Yes, I already commented that above. :-)

Juho