[EM] Fwd: U/P voting: new name for simple 3-level method.

[Kevin Venzke]

Hi Jameson, 

________________________________ 
De : Jameson Quinn <jameson.quinn at gmail.com> 
À : Kevin Venzke <stepjak at yahoo.fr> 
Cc : EM <election-methods at lists.electorama.com> 
Envoyé le : Vendredi 9 septembre 2016 8h38 
Objet : Re: [EM] Fwd: U/P voting: new name for simple 3-level method. 
>>> 
>>>Using parallel language to yours, here's U/P: 
>>> 
>>>If there's a majority acceptable (minority unacceptable), elect the one of those that's most preferred. Else elect the most 
>>>preferred. 
>>> 
>>>That's 4 words more complicated. You could do it in fewer than 4 extra words if you really wanted. 
>> 
>>Ok. Is it better than MCA? 
>> 
> 
>Yes. MCA is significantly more subject to chicken truncation strategy. 

Hmm. In MCA, from the perspective of a (potentially) majority faction split between two candidates, MCA is identical to Approval, 
so I understand that much. 

> 
>In discussion with C. Benham, I have realized that U/P is actually more similar to MTA than I'd realized. In the terms above, MTA 
>(Majority Top Approval) is: 
> 
>If there's a majority acceptable (minority unacceptable), elect the one of those that's most preferred. Else elect the most acceptable. 

Yes I mentioned this method but didn't call it MTA. I've called it MAFP outside of a ratings context. I think my wording shows how 
we might hope it would be used: 
"There is another method where you elect the approval winner unless there are multiple candidates with a majority, in which case 
preferred ratings break the "tie."" 

So, you use the middle slot when you calculate it's more likely you need to "vote against" that candidate in the tie, than "vote 
for" him. That is a murky statement (because you do not know who is in the tie), but simple enough that I don't dislike MTA that 
much. Functionally though I'm not sure how different it is from Approval. 

>So, U/P and MTA differ only when all candidates are majority unacceptable AND the most preferred differs from the most acceptable. I 
>think that this situation is on the whole unlikely, and that when it occurs, it will probably be due to strategy. Furthermore, it may 
>enter into consideration as a contrafactual, in which case it is by definition strategically motivated. 
> 
>I think MTA is more likely to get it "right" if the ballots are unstrategic; but that U/P is more likely to get it "right" if the 
>ballots are strategic. Since, as I said, I think the latter is more likely, I think U/P is better. Their simplicity is identical. 

Hmmmm. I would think and hope that voters would use strategy under MTA, so of course any resulting ballot set would be "due to 
strategy." I don't think I would characterize U/P differently, though I think the strategy is harder to articulate. 

It sounds like you're saying if this situation on the ballots arose, it would probably be "due to strategy" (regardless of which 
method's strategies created it), and that U/P would resolve it better if it is due to strategy (i.e. U/P's strategy). 

What do you make of these ballots: 

43: A 
24: B>C 
23: C>B 
10: D 

Under MTA the B and C voters are being completely reasonable: They hope for majority approval but can still hope for a win if they 
don't get it. 

Strategy is less likely to produce these ballots under U/P because the B and C voters are taking a gamble. To get a similar outcome 
they have to vote B=C. Anyone who doesn't is functionally defecting! 

> 
>But I don't want to be "right" if it leads to me just arguing in my own corner. I'd far rather choose a system that can get consensus 
>from as many theorists and activists as possible. That's why I continue to put approval forward as the first reform option. I'm only 
>working on discussing a three-slot option in order to have a back-up suggestion for people who object to approval because of the issues 
>of non-expressive compromise and/or chicken strategy (including concerns over spoilers; I think that in approval, "spoiler" almost 
>always refers to a chicken-type scenario.) 
> 
>So I'd be happy to give up on U/P and embrace MTA if that would help build a clearly broader consensus. 
> 
>(One side advantage of U/P is that googling "MTA" is never going to send you to a voting page.) 

I see. Yes, I have doubts about advocating Approval because it doesn't seem like a strict improvement, even if on balance it should 
be a large "net" improvement. 

> 
>>>>So you are saying if nobody manages to get majority approval, you will only be using the top ratings and ignoring "acceptable" 
>>>>ratings? 
>>>That is unusual; my instinct is that if we can't find a majority we should try to find votes to get as close as possible. There's 
>>>a 
>>>risk that you are collecting enough information to permit concluding e.g. that a simultaneous approval and Condorcet winner lost. 
>>> 
>>>Yes, this is possible. But consider the kind of scenario where it happens: 
>>> 
>>>20: A>B 
>>>20: A,B 
>>>19: B 
>>>2: C>B 
>>>39: C 
>>> 
>>>In this case A wins, even though B is approval and Condorcet winner. But I find this kind of thing very implausible in practice; 
>and 
>>>if such a pathology occurred, it would not be a horrible outcome. It's implausible because the electorate above is bizarrely top-heavy 
>>>in its ratings; except for the C>B voters, who could make B win by voting C,B, in view of the fact that C is a clear loser against A 
>>>and B. 
>> 
>>This specific thing is probably unlikely, but my general concern is that you might collect a bunch of middle slot ratings without 
>>doing anything with them, and public might be able to speculate about whether it made sense. 
>> 
>>I am not sure why you called the above scenario bizarrely top-heavy. It's a top-heavy method. I'm not too clear on when one should 
>>be using the middle slot really. 
> 
>It's a top-heavy method in that it focuses on top ratings. That does not mean, I think, that it would lead to top-heavy ballots in which 
>voters tended to rate several candidates as preferred. 

Well there are only 20% "anyone but C" voters here. They seem to be voting pretty reasonably to me. 

>Usually that's the point of "top two", but not in this case. In this case, it's just a simple heuristic for "is there some 
>non-dark-horse candidate without majority unacceptable?" You could get the same effect in other ways: 
> 
>- A quorum rule: non-votes for X count as "acceptable" iff over 50% of voters have rated X explicitly. 
>- A partial rule: non-votes count as 0.55 of an "unacceptable" vote and 0.45 of an "acceptable" vote. That way, a dark horse would 
>need "prefer" to beat "unacceptable" by about 9% to win. 
> 
>I'd consider either of the above to be fine, but my guess is that the "top two" rule I've already enunciated is the most politically 
>feasible. I think rules like the two suggestions just above seem more arbitrary and strange. 
> 
>>>>it feels a little schizophrenic to me that you really want winners to have majority non-disapproval yet do not actually think it 
>>>>indicates a worthy candidate. 
>>> 
>>>I think that majority explicit non-disapproval does indicate a worthy candidate; and that majority implicit non-disapproval is 
>>>better than majority disapproval IF the candidate achieved sufficient scrutiny (which we can assume is true for the two frontrunners). 
>>>But I think that unless we have evidence that a candidate got scrutinized, majority of combined unacceptable or didn't-bother-rating 
>>>is as bad as a majority of unacceptable. 
>> 
>>Sure, I just wish it seemed more organic. 
>> 
> 
>If you have a better idea, please share. I wouldn't be surprised if we came up with something better. 

I will keep it in mind, but I think you would make a model of some kind, that would say what makes us think that a given candidate got 
scrutinized or not, and then "degree of scrutiny" would be part of a single formula. 

>>>> 
>>>>I tend to think we will be lucky if we can consistently get even one majority non-disapproved candidate in elections. 
>>> 
>>>I disagree. The people who hate both frontrunners in the current system are louder than those who are OK with either; and in the 
>>>current presidential election, they might well be more numerous; but in the average election, I think that "either one is acceptable" 
>>>is more common than we might realize. 
>> 
>>Hmm... Voters who are OK with either? I have a lot of questions about that. Firstly do these voters really exist. And do they vote? 
>>Would they be willing to put on the ballot that they are OK with either? 
>> 
> 
>Remember, this starts to matter in 3-way elections. The closest cases like that in US presidential history are '92 and '96, with 
>Perot. I think that there were probably a non-trivial minority of "anybody but Perot" voters; that is, people who would have voted 
>something like Bush>Clinton>Perot or Clinton>Bush>Perot in U/P. If those people outnumber the Perot>...>Clinton,Bush people, then 
>it's likely (guaranteed?) that one of Clinton or Bush would get a majority acceptable. 

Are we still talking about "the average election" at this point? You seem to say that a voter could be OK with both frontrunners 
as long as there is, actually, a *third* contender. (If you say Perot does not need to be so strong as that, then I don't think 
I would agree that there would be a lot of people who would actually vote that the two frontrunners are OK as long as they are 
not this weaker candidate.) 

Kevin