[EM] [CES #9013] Re: EMAV?

[Abd ul-Rahman Lomax]

At 12:32 PM 7/1/2013, Jameson Quinn wrote:
>I responded with a new subject header because I was still hoping 
>that Abd would respond to my earlier post, copied below:

I'll answer here.

>Abd:
>
>Frankly, I'm a bit frustrated.

My condolences. Have you tried breathing exercises? I recommend 
recognizing frustration and the source of it and dropping the story 
involved. But, of course, you may think I've done something Wrong. Do you/

>One of the main reasons I proposed MAV in the first place was that 
>you seemed to support it.

I looked at an idea and realized that it would resolve a certain 
issue, a certain kind of LnH failure. I thought, then, that MAV might 
be a decent compromise, if needed to get over the Later-no-Harm 
objections that FairVote will surely raise.

However, it does that by sacrificing social utility, and that's 
clear. Finding a multiple majority at a lower rank is likely to be 
reasonably common in contested public elections.

>  You've done a good job expressing the advantages of Bucklin 
> systems, and I wanted to get past the point where I was talking 
> about GMJ and you were talking about "Bucklin" as an individual 
> system, since the differences are trivial.

Right. Now, we could propose simple Bucklin. It is reasonable from 
the point of view of "instant runoff approval." That is exactly what 
Bucklin is. But MAV complicated simple Bucklin, and it tosses out the 
votes that create the multiple majority, instead resolving the 
election on the basis of the round prior to that. Now, the good news 
is that it selects the winner from the multple-approved set, so the 
winner *is* approved, still, but I started to wonder why we were 
effectively tossing the score data, that would show the true social 
utility optimizer.

And then I realized that we could combine Bucklin with Range, where 
the election resolves normally as Bucklin, that is, collapsing 
approval, but then uses the Range data from the ballots where this 
leads to certain conflicts.


>I still feel that way. Sure, I have an opinion on whether MAV or 
>EMAV is better (I prefer MAV).

Obviously, I disagree.

>  But the similarities are more important than the differences. 
> Unless we're all willing to give up on getting our own personal 
> ideal in every detail, we're never going to have enough unity to 
> make a difference.

Great, Jameson, what are you proposing to give up? Actually, I *would 
not want you to give up anything important.* I would *much rather* 
seek an underlying consensus and build back from there.

We are tweaking Bucklin, with MAV, *away from Range.* It can be 
expected to damage Social Utility. There should be a good reason for 
that, don't you think?

EMAV tweaks it *toward* Range and actually uses Range amalgamation, 
much more refined and sophisticated than mere Approval.

>To put this in stark terms: I think it's far more of a danger that 
>voting reform will fail because of the picky disunity of people 
>proposing their personal best system, than that it will pass but 
>lead to some bad result because of a minor flaw in the system.

So, because of a few days of delay, which could become a few weeks or 
a few months, we will suffer this Failure to Agree, and the whole 
voting reform will go south?

I suggest considering the underlying principles here. What makes one 
voting system better than another? What is our overall goal as CES?

>And it's clear that you do see the importance of the similarities, 
>or you wouldn't have dubbed your proposal EMAV.

No. I did not pick EMAV as a derivation of MAV. I picked it as 
descriptive. There *are* similarities, substantial ones. However, 
what's the difference? Does the difference improve or damage the 
system performance or its implementation possibilities?

What is a bit complicated about EMAV is the range ballot, i.e, the 
values assigned to votes. I think I just wrote that those could be 
given values which clear discriminate between approvals and 
disapprovals: 2, 1, 0, -1, -2. In fact, though, on the principle of 
range being fractional voting, it is 1, 0.75, 0.50, 0.25, 0. 
Explaining this ballot is explaining a range ballot.

The range ballot is then used to control voting in a series of three 
approval elections, followed by, if those fail to find a majority, a 
range summation of the values for a result, or for runoff nominations.

The concern about later-no-harm is handled as it is in range: the 
lower preferences do not receive full strength. This differs from 
MAV, which *discards* the lower preferences if a multiple majority is 
found, or, if there is full amalgamation, it gives the lower 
preferences *full approval value.*

EMAV balances two goals: SU optimization and the seeking of majority 
approval. MAV does only the latter, using a coarse approximation for 
the former. Why not use the real range data?

>So I'm willing to have a discussion about the relative advantages 
>and disadvantages of MAV versus EMAV.

Thanks.

>  And I'm willing to commit to following the majority consensus of 
> that discussion; if, at the end of that discussion, there are more 
> voices in support of EMAV, then I will use EMAV as my default 
> second proposal (after approval).

That is not how consensus works. It's not a vote. Jameson, you are 
free to do what you choose, but if we are going to have a vote, I'd 
want to do it much more formally and with much more preparation.

>  But I haven't yet heard a similar commitment from you: when the 
> topic is what comes after approval, will you agree to try to show 
> unity in what system you mention first, rather than just rambling 
> about whatever occurred to you most recently?

Do I get that you don't like my conversational style?

>I'm sorry for the harsh language, but you should understand that 
>focus is a recurring issue with you. Politely addressing it only in 
>its specific manifestations seems not to work, so I'm consciously 
>deciding to be more forceful here. I hope you realize that I would 
>not be doing this if I didn't truly respect your intelligence and insight.

Sure. Just understand the story of the goose and the golden egg. 
Suppose that EMAV is actually the golden egg. We found it by these 
"rambling discussions."

Understand, as well, that I'm not writing everything that I think on 
this topic. I'm only addressing very specific issues that come up. 
More will be revealed.