[EM] General strategy and proposabiliy

[MIKE OSSIPOFF]

Replying to IRvists is a waste of time.

Regardless of what anyone says, they're going to keep repeating their confusions and lies.  The only time to answer them would be in a public voting-system enactment campaign, where your reply would be for the benefit of the general public.  Until then, there’s no point, and it’s just a waste of time.

But, I’ll just answer two of the IRV-promotional fallacies expressed by Dave. But only this one time. 

Dave said that Approval can be “gamed”. What does that mean? (That’s a rhetorical question, Dave. You needn’t answer it, and won’t be replied to if you do.)

Gibbard & Satterthwaite showed that all nonprobabilistic ballots-only voting systems have strategy. Also, that’s obvious from our experience with this subject.

Not only can IRV be “gamed”, but it _must_ be gamed, to avoid a Burlington. We’ve been telling the IRVists this for decades, though in former decades, it wasn’t called a “Burlington”.

And we’re talking about an especially drastic “gaming” need: Favorite-betrayal.

Not only is IRV’s strategy more drastic than that of Approval,  but it’s also  more complicated and less known.

I suggest that, in our public political elections, there are completely unacceptable candidates who could win. I call that a u/a election. u/a stands for “unacceptable/acceptable”.

So, first, say it’s a u/a election.

Approval’s u/a strategy is unmatched in its simplicity: Vote for all of the acceptables, and for none of the unacceptables.

Approval’s power and straightforwardness are unequalled.

IRV? Rank the acceptable candidates in order of their ability to take the election away from an unacceptable. Of course you  could then rank, below them,  the unacceptables in order of preference, but there’s little if any reason to do so, if they’re all unacceptable. 

Of course, that ability to take the win away from an unacceptable is a combination of several abilities:

To get enough initial votes to escape elimination till more votes arrive as transfers
To get transfers soon enough, and in sufficient amoun,t to avoid elimination throughout the count.
To get a majority, or to remain after all others are eliminated (which of course results in majority too).

What about non u/a elections?

We’ve extensively discussed Approval strategies for non u/a elections.

For IRV, no one knows. It would be a matter of ranking the candidates in order of some unknown combination of a) (some unknown measure of) winnability; and b) merit.

And don’t try to say that, because IRV strategy is unknown, you can just rank sincerely. Remember Burlington? No, you’d better top-rank your needed compromise, even at the price of burying your favorite. It’s just that you’ll be doing so without any clear guidelines. 

Dave also said something to the effect that, it’s hopeless to get enactment when there are so many methods to choose from. So (his argument goes) we’d better propose IRV, rather than having to choose among the others.  :-)

But that is a problem of rank methods: There are innumerable ways to count ranked ballots. So, whichever rank method you propose, people could ask, “Why _that_ rank-count?”

The IRVist creed seems to be that, because someone could afford to push IRV through in some local jurisdictions, then IRV is the thing to go with, because it has that start. But need I repeat how that start is less than flattering for IRV? IRV is a heavily-financed loser. And even Richie’s money won’t be able to buy that loser’s way into enactment for federal elections.

So, the already-had experience with IRV counts against IRV more than for it.

So what’s the answer to “With so many methods, why should we choose the one you propose?” How can we avoid that problem?

The answer is obvious: Propose the simplest, most straightforward and natural change from Plurality that will get rid of Plurality’s artificial favorite-burial problem: Approval. Approval is simply Plurality done right. 

Approval has the co-operation/defection problem demonstrated by the Approval bad-example. That problem isn’t strategically insuperable, even in unenhanced Approval. As I said a few days ago, Forest suggested a solution in the form of a diplomatic offer that the A voters could make to the B voters. Previously, I had suggested a publicized promise of principled refusal of an unacceptable compromise. 

I’d also pointed out that successful defection by B voters would have consequences in subsequent elections. It would lose the support of the A voters, and that would be, in the long-run, a mistake for the B voters. In any case, over a few elections, experienced, trial & error, has a way of dealing with such problems.

In fact, as I said a few days ago, it’s been shown by Myerson & Weber that Approval will soon home in on the voter median. That was also demonstrated on EM, some time ago.

…in stark contrast to Plurality and IRV, which, they showed, can continue to elect the two most despised parties forever, at Myerson-Weber equilibrium. (in which the results (one of “the 2 choices” always wins and no one else gets any votes to speak of) seem to confirm the belief that those 2 parties are indeed the 2 choices).

But it’s possible to avoid the co-operation/defection problem, in an obvious, natural, straightforward way, with an obvious enhancement of Approval: The conditional methods, starting with Optionally-Conditional Approval (AOC). 

Less important, the next kind of straightforward improvement would be to increase the number of levels of majority rule protection, by such obvious transitions as the transition from AOC to MTAOC or MCAOC.  Or even AOCBucklin.

I’ve suggested conditionality by mutuality, but these methods could also use conditionality by top-count, where a conditional middle rating is actually given only if its recipient has more top-ratings than any of the voters’s top-rated candidates.

Though not quite as straightforward as AOC and MTAOC, other approaches to conditionality by mutuality are MMT and GMAT. MMT’s mutual majority set, and GMAT’s mutual approval set are natural ways to achieve conditionality by mutuality.

I emphasize that all of these are natural, obvious and straightforward enhancements of Approval, to get rid of its co-operaton/defection problem, and (less important) to gain additional levels of majority-rule protection.

That naturalness, obviousness and simplicity, of Approval, and its various derived methods named above, is the answer to the questsion “How can we justify proposing some particular method, when there are so many?”

The fact that these Approval enhancements can be offered as _options_ in an Approval election further increases their proposability and makes them more difficult for opponents to object to.

As I said yesterday, the conditional methods avoid the strategy problems of the other nonprobabilistic ballots-only methods, without adding any problems. 

For nonprobabilistsic ballots-only methods, there’s no such thing as getting rid of strategy, but the conditional methods show that it’s possible to get rid of the genuinely problematic strategy problems.

Mike Ossipoff

 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20120126/d1e38cfb/attachment-0002.htm>