[EM] Criteria (Re: IRV ballot pile count (proof of closed form))

[Abd ul-Rahman Lomax]

At 03:34 AM 2/6/2010, Kristofer Munsterhjelm wrote:
>Abd ul-Rahman Lomax wrote:
>
>>I don't think it's true that it has been "without problems." There 
>>are and have been problems. But if IRV were an optimal method, it 
>>might be worth the trouble. For multiwinner STV, indeed, it might 
>>well be worth the trouble. But for single-winner? I don't think so. 
>>There are simpler methods that produce better results, by all 
>>objective measures.
>>(Frankly, there is only one clearly objective measure, which is how 
>>a method performs in simulations, particularly with reasonable 
>>simulation of actual preference profiles -- full utility profiles 
>>-- and voting strategies as voters are known to use or are likely 
>>to use. "Election criteria," like the Condorcet Criterion, tend to 
>>be criteria that are intuitively satisfying, but that can actually 
>>fail completely and obviously under certain conditions, and a 
>>method failing a criterion may mean nothing if the failure is so 
>>rare and requires such unusual voting patterns that it will never 
>>be encountered under realistic conditions. Basically, how do we 
>>judge the criteria? And there are only two ways that I see, one is 
>>through utility analysis and the other through basic democratic 
>>principles, broadly accepted, such as the right of decision that is 
>>held by a majority; a majority of voters voting for a single 
>>proposition, with no opposing majority voting simultaneously for a 
>>conflicting proposition, must have the right to implementation. 
>>When there are multiple majorities there is not a simple question 
>>and there remains doubt as to a majority decision.)
>
>I'm not quite sure about this. Say you have an almost-perfect 
>method: usually it elects great candidates, but once in a while, it 
>picks a very bad dictator, or respects the wishes of a tiny 
>minority, or somesuch. It does this seldomly enough that it's just 
>ever so slightly better than the best alternative, on average.

Fine. I don't see this as a response to what I stated.

>However, those who hold democracy as an ideal would probably not 
>like this method, because once in a while, it "hiccups". The jitter 
>or hunting, itself, provides a bad outcome; and in a sense, criteria 
>are guarantees that a bad outcome (according to the criterion) won't 
>happen, period.

Right. Bad outcome according to a criterion. But an outcome that 
satisfies the criterion can cause a revolution, it's so bad. The 
point is that the Condorcet and Majority criteria *can* fail in this 
way. It's very rare, as to anything so major. Normally a Range result 
and the Condorcet winner coincide. But if they don't, it means that a 
minority with a large preference is faced with a majority with a 
small preference.

Normally, the difference isn't great. Or maybe it's just an artifact, 
i.e., maybe the majority actually did have as strong a preference as 
the minority, but for some reason didn't vote that way. But what if 
it is real and what if it is significant, and what if the majority 
margin is very small? What this means is that the minority, which is 
roughly as large as the majority, has the incentive to defy the 
election result, and has the motivation (that's what preference 
strength means, really, it means that they will invest in it, if needed).

Note, however, that I have not recommended awarding an election to a 
range winner over a majority or condorcet winner. I've said that this 
is an ambiguous situation and should be tested with a runoff, unless 
election data can conclusively resolve it (by predicting accurately, 
based on prior experience, not just theory) what the runoff result will be.

My position is that the majority has the right of decision, it's a 
general democratic principle. But it does this through binary 
decisions, in standard deliberative democracy, never through 
multiple-choice elections, unless they show a majority result. 
However, once we have a voting system that collects better data, it 
becomes possible to detect a range anomaly in a majority result, and 
to ask again, with the electorate now informed from the first 
election and an opportunity for the minority to explain itself in a 
more focused environment.

It's still up to the majority. I've argued, though, that in a runoff 
between a range winner and a majority winner, the range winner, 
assuming that the range gap is real, has a leg up, because of effects 
on turnout. If it is true that the majority has low preference, they 
will not turn out to vote in as high numbers as the 
stronger-preference minority. And this is classic; quite simply, it's 
not uncommon that, for some reason, a majority result is 
reconsidered, it goes the other way. I read an example the other day 
in the history of the Alcoholics Anonymous World Service Conference, 
that a measure approving a publication passed by almost-unanimity. 
AA, however, always asks the minority if they want to present a 
report. They did, and someone moved reconsideration, apparently, and 
the new vote rejected the same publication with almost unanimity.

What was the first preference of a majority there?

On the other hand, if the range anomaly was an artifact, the result 
of poor strategy -- or effective strategy on the part of the 
minority, perhaps -- then the runoff will show that, there will not 
be the turnout differential and the majority position will be sustained.

When people care they turn out to vote. Not caring can mean despair, 
or it can mean satisfaction with either outcome. It's important, I'd 
say, for every vote to count, and taking this to an extreme, Asset 
Voting. Every vote is not only counted, it is *effective*, and the 
effect is visible and trackable.

>To make it somewhat more familiar: Range may be the best Bayesian 
>Regret method, but that won't help once people notice that it gives 
>a minority power to outvote a majority.

But I'm not recommending that at all. What Range/Bucklin would do is 
to give a minority the power to, under some conditions, call 
attention of the majority to an anomalous situation requiring a 
little more attention, asking for a reconsideration.

And this is only as an advanced reform. Much more important is 
getting majority results, which Bucklin does a great job of doing, I 
expect, if used in runoff voting. It would avoid many runoffs, but 
not all. Used in a runoff, it would allow write-in candidacies that 
would not spoil elections (this is important in California, where 
write-ins are allowed, by default, in all "elections," and that has 
included runoffs unless specifically excluded by local law. Used to 
include runoffs until San Francisco figured out it could stop a 
serious runoff challenger and outlawed write-ins in the runoff, based 
on a supposed desirability of getting a majority. Right. Get a 
majority by preventing someone from running, that's cart before the 
horse! And the California Supreme Court, against all parliamentary 
precedent, decided that a runoff was part of the same election, and 
if write-ins were allowed in the primary, that was enough. And, 
folks, we were asleep at the switch, because we had our heads buried 
in the ideal single-ballot sand, and forgot to defend the most 
prevalent election reform, runoff voting.

And, of course, FairVote has been picking it off. They even named the 
method IRV as a way to promote it, to go after vulnerable runoffs.

>  Sure, that may be "better" according to BR, but it's not 
> majority-rule democracy, which is the context in which these 
> methods are considered. If you're going to fail Majority, you at 
> least need a runoff so it's intuitively possible for people to keep 
> that from happening.

Bingo! In other words, overall, we are not going to fail Majority. 
Twere it up to me, we'd *require* a majority, and that's practical 
with Asset Voting. It's done in parliamentary systems! They don't 
elect the prime minister by plurality, period.