[EM] Clean Government Alliance

[Kristofer Munsterhjelm]

On 01/18/2013 06:46 PM, Richard Fobes wrote:
> On 1/17/2013 10:49 PM, Kristofer Munsterhjelm wrote:
>>
>> The general pattern I was trying to think of, in any case, was this: the
>> society is too far in one direction (according to the people). Candidate
>> X has a position solidly on the other side and brings the policies in
>> that direction. As X pushes policies towards the center, he gains
>> reputation for doing something well. Then as X goes past the center, the
>> people think "we'll give him some time; he's been right in the past, why
>> shouldn't he still know what he's doing?" And so it takes time before
>> the people recognize how far off the other side X really wants to go.
>>
>> Term limits mitigate this ...
>>
>> I have also been reading about predictor or ensemble systems (like
>> weighted majority voting). In that context, it's like an expert that
>> tends to be very right, but situations change and he suddenly stops
>> being right. It then takes some time for his weight to be reduced,
>> because he has such a high weight already. In dynamic situations (where
>> experts may often shift from being very good to not being good at all),
>> sliding window versions of WMV (or UCB) do better than non-sliding
>> versions. I can find papers for this if you're interested :-)
>
> Currently, in politics there is not a close correlation between voter
> preferences and who ends up in office, so the tendency you claim does
> tend to occur.
>
> However, if elections are improved so that there is a high correlation
> between voter preferences and who ends up in office, then such
> "over-runs" would quickly lead to a push back to center.
>
> Such over-runs are a component of the concept of "resonance" in Physics.
> This over-extended "state" quickly lead to an ever-increasing push back
> to center. Yet, overall, the result is an oscillation that averages out
> to be centered.

In my description, the problem is that the people trust the politician 
as he shifts from interests aligned with the people to interests not 
aligned with the people. They say "he's done right things in the past, 
so he knows what he's doing now, too". So the effect is one of people's 
judgement of the politician, rather than how that judgement is being 
distorted by the election method.

Do you think people are actually quick to react against overrun and the 
method is the problem, rather than the people's estimates?

I suppose one way to find out which is the case would be to check if 
such overruns have happened in PR countries. However, this might 
complicate the situation, because PR countries - at least parliamentary 
ones - don't have a Micawberian "50% + 1, I win, 50% - 1, I lose" 
situation, so the gradual feedback from the parts of the people that do 
change their opinions often might inform the parties currently in power 
that if they don't do something, more people will follow.

(I think this is part of the reason the Labor Party here is moving to 
the right. Their coalition will most likely lose their majority in the 
next election anyway, though.)

So it would look like the way to find out which is right would be to 
find a country with an advanced election method but not PR. But I don't 
know of any such countries since I don't consider IRV an advanced 
election method. One might stretch it by considering TTR advanced - it 
is certainly better than IRV in my eyes... so have there been any 
overruns in France? I'm not sure.

> If, after election-method reform, there should be a need to "dampen"
> such wild swings, there other -- and I believe wiser -- ways to do so.

Which methods or ways would you suggest?

My background, such as it is, regarding these problems is more a 
cybernetics and CS one rather than a Physics one - though I don't have 
any degrees in either and therefore work more on intuition than actual 
calculation.

In any case, I have been thinking about methods to dampen raw populism 
in more responsive democratic systems like Liquid Democracy. Here the 
problem (if there is one; we don't know for sure since there haven't 
been any public experiments with the thing that I know of) would be that 
the people react too quickly rather than too slowly. Now, a traditional 
response would be to say that "this is the problem of too frequent 
elections". To me, that seems to be like saying that a thermostat that 
regulates temperature and oscillates too wildly around the setpoint has 
a problem because it samples temperature too often. You *could* say 
that, but you could also just slow the response of the controller. It 
seems a waste to force less information to be gathered instead of more 
just so that the system should work. In a LD case, that might take the 
shape of using a statistical method with a high breakdown point instead 
of sum or average for the power calculations, or by gradually 
adding/removing the voting power given by a subscription rather than 
just going directly to 0% or 100%.

So I can definitely understand that more sophisticated methods could do 
better. I'm thus wondering which methods you had in mind.