[EM] Dave: More levels of expressable preference

[Dave Ketchum]

On Fri, 21 Oct 2005 20:11:05 +0000 MIKE OSSIPOFF wrote:

> You'd said:
> 
>> Can happen that I desire electing Tom and hate Harry. Further, I am 
>> not sure Tom can win so, in case he loses (and only if he loses), I 
>> want to do what I can to make sure Dick gets in as my second choice.
> 
> 
> You add:
> 
> Let me expand on this.  There is a race for mayor in a city near here, the
> "aroma" is unavoidable and, therefore, if I got to vote, my primary goal
> would be that Harry lose (and I do not trust the polls, for I suspect they
> work for Harry):
> 
>      Assume Tom and Dick are equal in my eyes.  Therefore I would be
> happy with Approval and any flavor of Condorcet that permitted =.  IRV is
> not too bad
> 
> I reply:
> 
> As proposed and used, IRV is no good. Unless you know which of {Tom, 
> Dick} is the more winnable one you need to beat Harry, then you have to 
> risk his getting eliminated immediately before your traveling vote gets 
> to him. Of course if {Tom, Dick} is a mutual majority, then IRV would 
> work ok in that particular instance. But, in general, mutual majorities 
> are a fortuitous special case.


INTERESTING:
      You write that IRV is "no good" because it cannot give equal ranks.
      Yet you demand that I accept Approval even though I object to its 
inability to give unequal ranks.

> 
> You continue:
> 
>      Or one of my two favorites earns a black eye - not enough to become
> a reject like Harry, but enough that I must do all I van to help ny
> remaining favorite.  Now IRV (except for spoilers) and Condorcet are 
> fine, Approval imposes an intolerable decision.
> 
> I reply:
> 
> No it doesn't. Is Dick acceptable or unacceptable? Vot for him if he's 
> acceptable. Don't vote for him if he's unacceptable. Yes, if Dick is 
> genuinely on the line between acceptability and unacceptability, then 
> you have to flip a coin in Approval (or give Dick 1/2 maximum points in RV).


Right back to our Approval disagreement.

Not having trustable polls, LIKING Tom and HATING Harry, I want Dick only 
in case I cannot have Tom - and want Dick if the only available 
alternative is Harry.

> 
> In any case, you know what to do.


Yes - curse whoever inflicted Approval on me when they should have known 
better.

> 
> If you know what you're doing, Approval is still ok, though not as 
> deluxe as one might like.
> 
> The serious problem happens when people don't know what they're doing, 
> when they're seriously mistaken about whether Dick is acceptable. Then 
> they can't be trusted to be able to use Approval in their own best 
> interest, and MDDA would be a much better thing to have. You'll get no 
> disagreement from me on that.
> 
> You continued:
> 
>      Or my falling favorite joins Harry for being unacceptable.  Now any
> method that tolerates bullet voting becomes acceptable, including
> Plurality and Approval.
> 
> I reply:
> 
> I agree.  Most any method is acceptable then, except for Borda, which in 
> its standard form, forces you to rank all the candidates.


I read elsewhere that Borda does not have this problem.

> 
> Ranking all the candidates in IRV may be a waste of time, but of course 
> it doesn't do any harm for the law to require completel rankings, other 
> than more time in the voting booth.


If my intent was bullet voting, I would choke if a variety of IRV which 
forbid truncation was inflicted on me.

> 
> You continue:
> 
> Thus Approval beats Plurality, but is not an acceptable destination.
> 
> Approval beats anything that doesn't meet FBC, for the purpose of our 
> public political elections.
> 
> In an early posting I said:
> 
> "I liken Approval to a solid, reliable handtool. But the trouble with a 
> handtool, compared to an automatic CW-finding machine like MDDA, is that 
> you have to know how to use the handtool, because it's manual. You have 
> to do it yourself.


Having a variety of handtools in a toolbox, that I can select from as 
desired, is GREAT.

Trouble is, I have not heard of anyone offering a choice of methods, such 
that I could select as I entered the voting booth.

> 
> " [for our existing electorate] The question of using MDDA vs using 
> Approval (or RV) is like the question of driving your child to school vs 
> giving the car-keys to your child and saying 'Drive yourself to school. 
> We'll find out how you do'."
> 
>> You quoted me:
> 
> 
>> But if Dick is unacceptable, then no, I won't give him an Approval 
>> vote, even if he is better than Harry.
> 
> 
> You reply:
> 
> You seem to consider existence of yellow to be an impossibility after 
> labeling red "unacceptable" and green "acceptable".
> 
> I reply:
> 
> No, it depends on the candidate lineup. If there are any completely 
> unacceptable candidates who might win, then it's what I call an 
> acceptable/unacceptable situation. There are effectively only two 
> preference levels. There is only one merit boundary that matters.
> 
> And yes, I believe that we have an a/ua situation in our public 
> political elections. In fact the giveway progressives believe that too, 
> though they disagree with me about who is acceptable. They think the 
> Democrats are acceptable. Hence the serious problem with Approval 
> (Unless the giveaway progressives will be able to stop voting for 
> Democrats when, in Approval, Nader outpolls the Republican.
> 
> Another way to describe an a/ua situation is: The candidates can be 
> divided into two sets such that the merit differences within the 2 sets 
> are negligible compared to the merit-difference betweent the 2 sets.
> 
> In Approval, vote for the acceptables. Approval is the perfect method in 
> an a/ua situation, if voters know what they're doing.


These paragraphs brag about Approval's ability to recognize red and green 
- pretending that its inability to recognize yellow does not matter.

> 
> In Plurality, vote for whichever acceptable candidate is most likely, 
> with your help, to be able to deny victory to the unacceptables.
> 
> I reply:
> 
> MANY of us can see more than two colors, and want ballots to comply 
> since it is possible.
> 
> I reply:
> 
> What country do you live in? It's entirely possible that, there, it 
> isn't an a/ua situation.


I presented a current example from a city in New York State near where I 
live.  Such trios of candidates can happen most any place that allows 
voting among more than two candidates.  That example is a general 
election, but such trios might be more common in primary elections.

> 
> Anyway, even if it is, of course I agree that, because voters don't seem 
> very quiified to vote in their own interest,  we'd be much better off 
> with MDDA, MDDB, or MDD,ER-Bucklin(whole) than with Approval. As I said, 
> you'll get no disagreement from me on that.


Voter education may always be a problem - a reason for keeping the method 
simple BUT able to serve major desires.

> 
> I'd said:
> 
>> I can't expect Plurality advocates to like the better voting systems.
>>
>> But if you're saying that they'd have an especially difficult time 
>> with Approval, then I disagree. Approval is just Plurality done right. 
>> Plurality without the completely unjustified rule that a voter may 
>> only give a point to one candidate, only show support for one candidate.
> 

Let's forget this exchange - does not exactly match my thinking, but not 
worth sorting out.

> 
> You reply:
> 
> Seems like you are not responding to what I said - that many of those 
> used to Plurality would see what you claim as acceptable and not see a 
> need to move farther.
> 
> I reply:
> 
> You're saying that if we enacted Approval or RV, people would be 
> satisfied with it and wouldn't want to get a better method. For one 
> thing, if Approval did as it could, and went right to the voter-median 
> position, and stayed there, what would there be for anyone to object to? 
> But if the giveaway progressives keep on giving an Approval vote to the 
> Democrat, after the vote totals have indicated that they no longer need 
> to, then even they will know that something is wrong, and that they'd be 
> better off with a good rank method. How right they'd be!
> 
> So if someone then began advocating MDDA, people would understand the 
> need for it, if Approval obviously was not working (because people 
> weren't qualified to use it).
> 
> I'd said:
> 
>> I observed someone, voting in a Condorcet poll, rank Dick over Tom, 
>> even though that person preferred Tom to Dick. That peron did that 
>> because they felt that Dick had a better chance of beating Harry.
>>
>> Did I tell that person that they had no need to do that in Condorclet? 
>> No, because that would have been an untrue statement. Condorcet 
>> doesn't guarantee that you won't gain by voting someone over your 
>> favorite. Approval does make that guarantee, known as FBC.
> 
> 
> BUT - Approval does not let me vote Tom preferred to Dick, and BOTH 
> preferred to Harry.
> 
> I reply:
> 
> No, but at least it doesn't make anyone feel a need to vote Dick _over_ 
> Tom. Thereby, Approval avoids the worst kind of strategy need.
> 
> You continue:
> 
> Tell me again how voters can improve their odds by voting against their 
> desires - at least other than responding to or cooperating with strategy 
> plots.
> 
> You want an example? Ok, here's the scenario. I'll show that Condorcet's 
> Criterion is incompatible with FBC.
> 
> You can also regard this as a descrption of how Condorcet's method, in 
> any form, can reward someone for voting someone over their favorite:
> 
> Consider two voters whose preference ordering is: B>A>C.
> 
> We're considering how those two people should vote when everyone else is 
> voting in some particular fixed configuration. Say the configuration of 
> other people's votes (sincere or otherwise) is such that if those two 
> voters vote sincerely, we'll get a circular tie in which C wins. 
> Obviously, given any rule for solving circular ties, I could write an 
> example in which C would win.
> 
> Say that, in that cycle, B beats A. But B doesn't beat A by very much. B 
> beats A by only 3 votes.
> 
> If A beat B, A would be the BeatsAll winnner, and would win, because 
> we're assuming that the method meets Condorcet's Criterion.
> 
> Those two voters could obviously change the pairwise result between A 
> and B if they voted differently. Let's say that no other pairwise race 
> is close enough for those two voters to influence the outcome of. That's 
> a reasonable assumption.
> 
> If those two voters changed their ballot to A=B>C, then B would only 
> beat A by one vote instead of three votes. I could write an example in 
> which C still wins, whether B beats A by 3 votes or 1 vote. Say it's so.
> 
> If those two voters changed their ballot to A>B>C, that would reverse 
> that pairwise defeat, and A would be BeatsAll winner, and would win. As 
> I said, that's because we're assuming that the method in use is one that 
> meets Condorcet's Criterion.
> 
> By voting A over their favorite, those two voters have improved the 
> outcome for themselves.


Thanks.

Hard to complain about getting EXACTLY what you ask for, even if it is not 
exactly what you desire.

Also true that, with this cycle, changing A>B>C to B>A>C would switch 
winner from A to C.

Problem has to be set up precisely, in or near a cycle, and hard to 
predict whether such wins or loses.

> 
> Therefore the method violates FBC. Since the method could be any 
> Condorcet Criterion method, all Condorcet Criterion methods fail FBC.
> 
> Condorcet's Criterion is incompatible with FBC.
> 
>> But, though it isn't insincere [to vote equally in Approval two 
>> candidates between whom you have a preference] , it is contrary to the 
>> voter's interest if s/he unnecessarily votes for someone who isn't 
>> really any better than the person s/he is trying to beat.
> 
> 
> You replied:
> 
> If he rated Kerry as ugly as Bush he would have no problem, even with 
> Plurality.
> 
> I reply:
> 
> Agreed of course.
> 
> You continued:
> 
> My topic here was that to SAY I liked Kerry better than Bush, under 
> Approval I had to SAY I liked him as well as Nader.
> 
> I reply:
> 
> Now you've put your finger on the problem: The problem is if you think 
> that you need to say that you like Kerry better than Bush. That's why 
> I'd much rather have MDDA, MDDB, or MDD,ER-Bucklin(whole) instead of 
> Approval. We're agreed on that. With those good rank methods you can 
> vote Bad over Worse without voting Bad equal to Best. Believe me, for 
> the sake of the country and the world,  I want you to have a good rank 
> method.
> 
>> I'd said:
>>
>>
>>> Of course hopefully they wouldn't keep doing that in Approval, when 
>>> Nader outpolls the Republican, or when they find out what a Democrat is.
>>>
> 
> You reply:
> 
> This is a distraction, for we often have lists of candidates where best, 
> tolerable, and ugly compete.
> 
> I reply:
> 
> So what? If you know that Best has just outpolled Ugliest, then you 
> should know that now you no longer have to vote for Ugly.
> 
> Or if you finally ask yourself why you want to vote for an Ugly, and 
> decide that you don't want to, that would solve the problem too. You 
> aren't going to vote for one of the two uglies again in 2008, are you?
> 
> You continued:
> 
> Further, we often do not have usable popularity ratings.
> 
> I reply:
> 
> That's for sure. We have continuous, consistant and insistant propaganda 
> telling people that they have only two "viable" choices. And when people 
> believe it, if appears self-fulfilling, since the vote totals indeed 
> seem to confirm the propaganda.
> 
> That's a property of Plurality. Any FBC-failing method can have that 
> problem too, under the same conditions.

Huh!  Bad popularity ratings can afflict elections regardless of election method.
> 
> You wrote:
> 
>> You're saying that not everyone perceives an acceptable/unacceptable 
>> situation. Maybe. But many do, and they'd do anything to keep an 
>> unacceptable from winning. Actually believing that Kerry was 
>> acceptable, many even voted for him in 2004, though he'd promised to 
>> continue and escalate the Iraq war.
>>
>> Though they no doubt perceived Nader, Kerry, and Bush as being in 3 
>> different merit-categories, they mistakenly believed that Kerry is in 
>> the category of candidates who deserve a vote.
> 
> 
> Our opinions on that particular race should not get in the way of 
> objective thinking.
> 
> I reply:
> 
> You watch someone make the same mistake over and over, and guess what? 
> You pretty much know that he'll make the same mistake next time. There's 
> nothing un-objective about that prediction. For the giveaway 
> progressives, each race is the same, and it's the same giveaway every time.
> 
> You continue:
> 
> Again, often the voter sees more than two ranks of candidates, and when 
> that happens the method needs to support saying so.
> 
> 
> I reply:
> 
> I doubt that there's anyone who wouldn't perceive more than two 
> preference levels, in any reasonable selection of candidates.
> 
> But those finer merit distinctions are unimportant if there is a set of 
> candidates who are completely unacceptable and who could win. Then, 
> whatever the method, a/ua strategy should be used. That strategy is 
> simpler in Approval than in any other method. The trouble is that we 
> don't all agree on what is acceptable. For that reason I'd much rather 
> have a MDDA, MDDB, or MDD,ER-Bucklin(whole) instead of Approval.
> 
> So yes,  a good rank-balloting method is a nice luxury. But when voters 
> are all confused about what is acceptable, or when voters are thoroughly 
> led-by-the-nose regarding who is viable, who is a serious candidate,  
> and whose policies are closer to what the public wants; and if they're 
> going to ignore it when Approval's vote totals indicate that 
> giveaway-strategy is no longer needed--then yes, a good rank method is a 
> necessity rather than a luxury.
> 
> We don't disagree on that.
> 
>> However, though N>K is their sincere preference, they could regret 
>> voting N>K if it resulted in victory for B, who would have lost if 
>> they'd voted K=N>B. My point was that MDDA isn't going to make people 
>> afraid to vote their sincere preferences.
> 
> 
> You reply:
> 
> Agreed that after counting the ballots we can often picture a different 
> outcome with a different vote.
> 
> I reply:
> 
> How true it is! But that isn't the problem. The problem is that, when a 
> voter perceives an a/ua situation, that voter will use whatever strategy 
> it takes to minimize the probability that an unacceptable will win. 
> Burying one's favorite by voting Kerry, and all the Democrats, over 
> one's favorite is the correct a/ua strategy if you think the Democrats 
> are acceptable. That's why I want a method that meets FBC.
> 
> If we don't agree on what is acceptable, then there's going to be a  lot 
> of disagreement on how to vote, especially when the voting system 
> doesn't meet FBC. It's much worse when the method doesn't meet FBC.
> 
> 
> Assuming liking N>K and
>     Voting N>K gave a worse result than N=K,
>     Who makes a guarantee that a different collection of ballots could 
> not have awarded the win to Bush for voting N=K?
> 
> I reply:
> 
> It would be very rare for your vote to have any effect on the outcome. 
> Therefore, most likely your voting N=K isn't going to help Kerry, as 
> compared to voting N>K. But the probability is beside the point. The 
> fact is that, in methods that don't meet FBC, you have a better chance 
> of helping Kerry beat Bush if you insincerely move Kerry up to 1st 
> place, voting only Kerry in 1st place. In an a/ua situation, if you 
> think Kerry is acceptable and Bush is unacceptable, then you will 
> insincerely vote Kerry in 1st place, over Nader.
> 
> MDDA is much simpler than Condorcet, and is even simpler and more 
> briefly-defined than IRV. And MDDA meets FBC, SFC, and SDSC. There's no 
> need to accept any less in a rank method.

I do not have time to respond to this tonight.
> 
> Mike Ossipoff

-- 
  davek at clarityconnect.com    people.clarityconnect.com/webpages3/davek
  Dave Ketchum   108 Halstead Ave, Owego, NY  13827-1708   607-687-5026
            Do to no one what you would not want done to you.
                  If you want peace, work for justice.