[EM] Fwd: Question about "Voting System"

[Kristofer Munsterhjelm]

On 03/28/2015 08:42 PM, Alexander Praetorius wrote:
>
> ---------- Forwarded message ----------
> From: *Alexander Praetorius* <citizen at serapath.de
> <mailto:citizen at serapath.de>>
> Date: Fri, Mar 27, 2015 at 1:18 PM
> Subject: Re: [EM] Question about "Voting System"
> To: Kristofer Munsterhjelm <km_elmet at t-online.de
> <mailto:km_elmet at t-online.de>>
> Cc: Alexander Praetorius <citizen at serapath.de <mailto:citizen at serapath.de>>
>
>
> Thank you for your suggestion.
> I like the winsorized more than the clipped version.
> I don't get your argument for the "median", everything you showed was
> related to "mean".

The general argument is that if you want majority rule, it's possible to 
set up examples where any wider interval than only considering the 
median can violate majority rule whereas median always respects it. 
Intuitively, when you use something that is closer to the mean, you let 
the strength of preference matter, so if a minority feels strongly about 
an issue, they can override a majority that doesn't feel quite as 
strongly about it.

This might be what you want; I'm just saying that if you need to respect 
majority rule in every situation, you can't use anything wider than the 
median. My example was a way of showing the intuition behind that :-)

> So I just wonder, what kind of method could be used to enable the
> population to set the  "winsorized cutoff point" dynamically :-)
> Basically, it is about people's trust in their fellow voters and whether
> they believe they vote honest or not.
> The more they distrust, the higher they would set the cutoff point, the
> more relaxed they are in believing their fellow voters to express their
> honest, but maybe extreme opinion, the more they would maybe allow for a
> low cutoff point.

I don't think a method can determine the cutoff point on its own. 
Remember that it can't read minds: it only has access to the votes 
themselves. So if it is to adjust the cutoff, it has to do so based on 
whether the voters think it's a reasonable cutoff or not.

I can think of two ways to do this. Call them implicit and explicit.

The implicit suggestion turns the continuous vote into an advisory vote. 
That is, the method is used to find what most people would support, but 
isn't binding in itself. Then, after one has done that, there's a formal 
vote which may simply be of the form: "Do you want this proposal or 
change to take effect? Yes or no", where a majority (or supermajority) 
wins. One can then connect the system to an optimization algorithm that 
chooses the cutoff so as to maximize the chance that the motion does 
pass. If there is little strategy going on, this will happen when 
there's not much of a cutoff (because the more honest information it can 
make use of, the better its idea of whether it'll pass will be). If 
there is a lot of strategy going on, it'll happen when there's a high 
cutoff.

However, the idea is not perfect. First, it could equivocate: say there 
are some difficult issues and some not-difficult issues, and say that 
strategy is the main problem in the first whereas not having enough 
information from honest votes is the main problem for the second type. 
Then the algorithm can only pick one cutoff for all of these proposals. 
Second, it may oscillate or lag, as any dynamic system might - e.g. if 
there's a run of dishonest voting, it may push the cutoff too high for 
too long, and then only pull it back down later.

The explicit suggestion is more conventional. You set the parameter "by 
hand" (or rather, the voters do so) through a separate referendum. In 
that respect, it's analogous to amending or altering a constitution, and 
like altering a constitution, the vote should require a greater margin 
to pass than an ordinary vote. Ideally, it would also be a yes/no 
referendum, because having only two alternatives lessens the possibility 
for strategy.

One way to look at the reasoning behind the explicit suggestion is that 
the voting system is a method for the voters to come to an agreement 
about something without having to spend a lot of time performing costly 
deliberation (discussions in a council, etc). The voting system is then 
a tool that is used by the people, and when the people want to fix that 
tool, they should be able to. However, since the tool is being used as a 
way to find the voice of the people, the method that alters it should be 
constructed so that it can't be altered to favor a minority at the 
expense of everybody else - hence the more stringent requirements for an 
amendment as contrasted to an ordinary proposal.

(An interesting side thought: could Jobst and Forest's consensus methods 
be used for adjusting voting system metaparameters or for amending or 
altering constitutions?)