[EM] Why I think IRV isn't a serious alternative 2

[Abd ul-Rahman Lomax]

At 08:56 AM 12/30/2008, Kristofer Munsterhjelm wrote:

>I would say that the problem is not just that culture expects 
>alienation, but that a full on "everybody discusses with everybody 
>else" scales very poorly (worst case quadratic) so that the common 
>opinion never converges, or converges very slowly.

Yes, *of course!* This is precisely the problem of scale in 
democracy. It is not a voting problem, per se, but a *deliberation* 
problem. And if voters don't deliberate, how can they intelligently vote?

What I hit upon was a finesse: they vote for what they understand or 
at least believe that they understand. They entrust the rest to 
representatives. Who decides if the voter is competent to vote on a topic?

*The voter.*

Democracy is properly rooted, not in popular decision, as such, but 
in popular *consent* to the decision process.

>  This is somewhat related to Parkinson's coefficient of 
> inefficiency - as the number of members in a committe grows, 
> subgroups form and there's no longer direct discussion.

Yes; however, there is a classical response: when the group size 
reaches some point, subgroups form *and independently deliberate 
within themselves, then deliberate collectively through 
representatives; often the representation is totally informal, often 
proxy voting isn't allowed, *but when organizations seek consensus,* 
as many do, voting is not so important though when the group size 
increases beyond a few, polling remains useful to sense and detect 
the degree of consensus that has been found. I've seen Approval be 
very useful in this situation, I've never seen Range used, but it 
should be even more useful (but very small groups still may find it 
more cumbersome than is appropriate).

In the FA/DP model, proxies and their clients and subclients form 
what we've called "natural caucuses." The clients of a proxy may be 
connected through devices such as mailing lists or other tools, or 
even meet face-to-face when an organization is local, and may 
deliberate under the "supervision" of the proxy. The purpose of that 
deliberation is to advise the proxy, as well as for the proxy to 
communicate to his or her clients the reasoning behind positions, 
questions to be addressed, etc.

However, this takes us quite far from out topic. The relevance here 
was an approach to the study of voting systems to understand how they 
simulate -- or do not simulate -- direct democratic process.

Where the positions of the electorate are considered to be fixed -- a 
drastic assumption, actually -- then voting systems can probably do a 
pretty good job of predicting what a negotiation would produce as a 
maximally acceptable compromise, one which would be approved by the 
largest majority. (Multiple majorities are probably pretty rare once 
all the factors are broadly understood and the question reduced to a 
minimal one.)

>The networks of connections would presumably make groups that 
>readjust and form in different configurations according to the 
>political positions of the people - like water, hence *Liquid* 
>Democracy. I think vote buying would be a problem with that concept, 
>though, because if the network of connections is public, then those 
>who want to influence the system can easily check whether the 
>members are upholding their ends of the "bargain" (votes for money)

The matter has been considered in some depth. The problem with vote 
buying under Liquid Democracy or Delegable Proxy, and especially in 
FA/DP, where actual power is not delegated, only representation in 
deliberation and an informal and rough measurement of broad consent, 
is that it would be extraordinarily expensive, for a transient 
benefit. So ... the bought proxy votes and advises corruptly. In 
public. However, this proxy, in a mature system, only directly 
represents a relatively small number of clients, whom the proxy also 
trusts. So the proxy *may* publicly advise, and privately hint or 
advise to his or her direct clients that they check it out.

It is in the interest of the clients that the proxy get the payment! 
*Especially if they tell him about it.* So, is the payment 
conditional not only on the vote or advice of the proxy, but rather 
on ultimate success?

If the former, the proxy gets to collect and the clients just decide 
to investigate for themselves, and they themselves have natural 
caucuses whom they advise, and so they say, "My proxy, whom I respect 
greatly, seems to have missed somethin on this issue, so I'm advising 
the contrary of his advice."

If the latter, the decision by the clients will be made on a 
different basis. Further, since bribery would be illegal, I assume 
that would continue (I'm not sure what you call it when you corruptly 
influence someone's advisor, but if it isn't illegal, it should be), 
the agreement is unenforceable. Would you agree? Knowing that the 
person offering the bribe was unethical and could simply deny 
knowledge of the bribe when payment was demanded?

No, bribery under a system like this wouldn't be impossible, but it 
would be far less possible than with systems where power is actually 
concentrated. Then it becomes possible to offer an individual "power 
node" enough money that they would keep quiet. I've never seen it be 
a promise of payment instead of a prepayment, when these things have 
come out. The famous image, practically a stereotype: wads of 
unexplained cash in the representative's freezer.

Remember, bribery must be efficient for the offering the corrupt 
payment. If it becomes necessary to bribe too many people, it is 
cheaper to focus on providing better products or service or lower 
prices to attract favorable decisions. If a company can get a 
contract worth billions of dollars, spending a few million on 
corruption of a few key players makes economic sense, possibly worth the risk.

This is one reason why I've proposed delegable proxy in the FA 
context: it is a very heavy protection. Once we see how it works in 
free associations, we will know much better what needs to be done 
when the process is public and might actually transfer power (other 
than the power to advise, which, while it can be very "powerful," is 
only so within fairly tight limits.)

>>>  Perhaps there would be if you just can't reach an agreement 
>>> ("okay, this has gone long enough, let's vote and get this over with").
>>Absolutely. And this happens all the time with direct democratic 
>>process. And those who have participated in it much usually don't 
>>take losing all that seriously, provided the rules have been 
>>followed. Under Robert's Rules, it takes a two-thirds majority to 
>>close debate and vote. Common respect usually allows all interested 
>>parties to speak before the question is called.
>>However, it takes a majority to decide a question, period, any 
>>question, not just an election. That's the bottom line for 
>>democracy. Taking less may *usually* work well enough, but it's 
>>risky. It can tear an organization apart, under some circumstances. 
>>That's why election by plurality is strongly discouraged in Robert's Rules.
>
>Given the above, maybe advanced methods would have their place as 
>figurative tiebreakers when one can't reach a majority by other 
>means. Say that the discussion/meeting goes on for a long time, and 
>a supermajority decides it's been long enough. If there are multiple 
>proposals, one could then have an election among those (law, no law, 
>law with amendment, law without rider, whatever). If there is no 
>method that's good enough to provide the majority certification you 
>seek, there could be a runoff afterwards - but I'll note that a 
>runoff doesn't magically produce majority support, since if one of 
>the runoff candidates/options is bad, most would obviously align 
>themselves with the other. The Le Pen situation would be a good 
>example of that. Just because Chirac got 82%, that doesn't mean that 
>Chirac is best, just that he's best in that one-on-one comparison.

That's right. In a deliberative environment, actually, this situation 
would be no problem at all. Under some rules, an election is actually 
complete until the clerk's report is accepted. So a majority can 
essentially cancel any election and propose -- and pass -- something 
else. It's rare that the election result is so bad as to be worth the 
trouble. In a Chirac/Le Pen situation, it would have been. There was 
a very strong preference, apparently, for Jospin.

So, imagine, instead of Top Two Runoff, the initial election is STV. 
Majority required. But it is Asset Voting. The candidates get to 
recast their votes. It would be quite a circus, I believe, but a very 
useful one. Some minor party leaders don't deserve their positions! 
And the voters who vote for them, seeing their incompetence and 
intransigence in public deliberation, would reconsider their votes! 
In any case, if the electoral process doesn't meet a certain 
deadline, finding a majority by then, then the *entire election* is redone.

Responsibility. Right now, minor party leaders really have little 
responsibility, they don't do much that would cause them to be held 
to account. Suppose Nader in 2000 could have decided the presidency, 
by awarding his votes to one of the two major candidates? What would 
he have done?

Whatever he chose, he would have been responsible for the results.

>>>In short, you'd have something like: for very small groups, the 
>>>cost of involving a voting method is too high compared to the 
>>>benefits. For intermediate groups, iteration works. For large 
>>>groups, voting is the right thing to do, because iteration is 
>>>expensive and may in any event lead to cycling because people 
>>>can't just share the nuances of their positions with a thousand 
>>>others, hive-mind style.
>>Right. However, there is Range Voting, which simulates negotiation, 
>>actually. If there are stages in it, it more accurately simulates 
>>negotiation. There are hybrid methods which address most of the 
>>concerns that I've seen raised. However, having two possible 
>>ballots taken rather than one is a *huge* step toward simulation of 
>>direct process, so large that I'd be reluctant to replace TTR with 
>>Range, unless it becomes Range/runoff. Robert's Rules notes as 
>>another problem with the preferential method they describe that 
>>voters cannot base their votes in subsequent process on the results 
>>of the first election.
>
>You say that Range simulates negotiation (because people vote 
>according to VNM utilities and all that). If that's true, why do we 
>need a runoff? Is it because you want a true majority?

Yes. Normally Range will indeed find the best result. But vNM 
utilities can conceal even fairly strong preferences, or exaggerate 
others, possibly based on bad public data. None of the utilities show 
insincere preferences (i.e., preference reversal) but the *actual* 
goal we would want to pursue is overall absolute utility, though that 
*is* modified by the principles of consent and fairness. (Imagine the 
situation if, by executing some small minority of people, some 
overall improvement in utility could be obtained. We still would not 
want to accept it -- in my opinion -- if the overall improvement were 
small. If we could save lives with it, more lives than it costs, 
well, isn't that the best argument for capital punishment? (Please, 
I'm not arguing such.))

The basic principle of democracy is the consent of a majority to 
collective action. That's not the ideal, that is the *foundation*. 
(The ideal is unanimity.)

So, when the Range result hasn't been accepted by a majority, it 
*may* be wise to verify it. When there is a pairwise winner over the 
Range winner, this has become much more advisable. In this case, we 
apparently have a small preference of a majority against a large 
preference of a minority. In this case, the majority should consent 
to its minor loss of utility or satisfaction. It shouldn't be 
automatic, though it is possible that the rules might except trivial 
loss routinely without verification.

The fact is that if the public is organized independently, so that 
the public votes with relative coherence, Plurality works fine! So 
does any voting system, pretty much! However, humans are frail, and 
fixed terms of office are, in my view, unnecessarily risky, even when 
good choices are made in the first place.

The central problem is the organization of the public for mutual 
advice and cooperation, *outside* of government, the full realization 
of Montesquieu's separation of the executive and the judiciary.