Talk:Majority judgment/Archive 1

Archive 1

Merge with Equivalent Article/s

This article should be merged somehow with range voting since it is essentially the same. Really it does not deserve to have its own article at all. It is not a separate topic, and it is not a particularly new or exotic idea so I don't know why there is a weird reference to Balinski as if this was a groundbreaking development.

At the very least the range voting article and the majority judgment article should link to each other in their "see also" sections.

Its not the same thing at all, this one use median point, The median point of 1 2 and 6 is the number 2, the average on range voting is 3. — Preceding unsigned comment added by 201.79.60.55 (talk) 20:29, 21 March 2016 (UTC)

Clarification please

Can somebody clarify what the following sentence mean: " Then, one copy of that grade is removed from each remaining candidate's list of grades"? As to "After removing 16% of the votes from each", please explain how that number is obtained.--Theorist2 (talk) 15:07, 29 June 2011 (UTC)

Clarified. Homunq (talk) 21:26, 29 June 2011 (UTC)

Look at all the grades received by two or more candidates who are initially tied because they have the same median grade. The winner will be the candidate who retains the highest median grade after one grade equal in value to that median is removed from each of the tied candidates until only one of these candidate still has such a median grade.

What do you think?

stevebosworth@hotmail.com Stephen C Bosworth (talk) 05:34, 22 March 2023 (UTC)

MJ and Condorcet

MJ is being counted as having a qualified fail for the Condorcet criterion, because it passes in a strong Nash equilibrium. But what does that mean? Clearly, with approval and Range, the strong Nash equilibrium means that you vote approval-style, that is, top or bottom rating for all serious candidates. But you should never have to do that for MJ; it's strategic overkill and it makes the ratings meaningless.

Basically, the Nash equilibrium in MJ means that anyone who prefers the CW to their principal rival(s) should give the CW a winning rating. That probably means "very good" to be safe, although in most cases "good" is good enough. That could be seen as a dishonest vote; if you're forced to think strategically in even this one case, you can't just honestly rate the candidates.

But there's another way to think about this, which could be compatible with absolute judgment. If candidate X is a CW, they must be rated above their rival Y's median rating by a significant number of voters. If I would like to rate both X and Y low - that is, X above Y but below Y's median rating - that probably means that I acknowledge as valid those other voters' reasons for rating X above Y. So, in "dishonestly" rating X as "good", I could be honestly saying "Even though X is only a fair candidate from my personal perspective, I acknowledge that X is a good candidate for me and my fellow voters"; or, in a similar vein "I believe that if X wins, it would be a good outcome of this election". Homunq (talk) 11:52, 4 July 2011 (UTC)

Typo in "Judgment"

What is this? "Judgment" instead of "Judgement" ?? — Preceding unsigned comment added by 78.48.15.131 (talk) 17:52, 3 December 2011 (UTC)

It's an acceptable alternate spelling, and the one used by the method's proposers. ˜˜˜˜

"primary sources" tag on criteria section

Aren't peer-reviewed sources reliable, even if the author without peer review wouldn't be? 200.49.190.24 (talk) 02:07, 14 January 2012 (UTC)

Given ranks in the example

In the example, the ranks the voters give to each of the cities are derived by first making a preference list from the distances and in a second step giving each city a unique rank. But, this is not reasonable since it breaks the idea of the system and prevents the readers from understanding what Majority Judgment is.

The idea is to vote for each candidate the rank he deserves, not to search for each rank exactly one candidate. So, the indirection over the preference list is misleading and should be avoided. Instead, the cities should be directly graded with a rank because of its distances, maybe normalized to the distance to the farthest city from the voters city:

distance (in relation to distance to farthest city)	rank
< 25%	excellent
25% - 50%	good
50% - 75%	fair
> 75%	poor

So, the ranks would be as follows (in brackets: distance in km), distances taken from http://www.geobytes.com/CityDistanceTool.htm

Voter from/ City Choice	Memphis	Nashville	Chattanooga	Knoxville	Median
Memphis	excellent	poor (311)	poor (428)	poor (554)	poor
Nashville	fair (311)	excellent	good (183)	good (255)	good
Chattanooga	poor (428)	fair (183)	excellent	good (158)	fair
Knoxville	poor (554)	poor (255)	good (158)	excellent	poor

-- Arno Nymus, 2012-03-05 19:20 — Preceding unsigned comment added by 77.23.79.151 (talk)

For for Majority Judgment to pass "Independence of Irrelevant Alternatives", as claimed in the article and by Balinski (he redefined the criteria for ratings), each voter would normalize not just to the alternatives in the given election, but to any range of alternatives the voter has experienced over time, or could conceive, or be nominated. According to the conceptual model proposed by Balinski, cities which are not in the current election affect the normalization for each voter. For example, the North-Eastern city of Kingsport, Tennessee. (Please enter replies in new section "IIA Compatibility of Tennessee Example") Filingpro (talk) 11:20, 18 March 2016 (UTC)

I agree with the idea, but actually I think the ratings for the middle cities should not be normalized, but absolute. So 1-175 would be "good", 176-350 would be "fair", and >351 would be "poor". Actually, in miles, the nice round intervals 1-100, 101-200, and 301+ give the same results. Like this:

Voter from/ City Choice	Memphis	Nashville	Chattanooga	Knoxville	Median
Memphis	excellent	poor (311)	poor (428)	poor (554)	poor
Nashville	fair (311)	excellent	fair (183)	fair (255)	good
Chattanooga	poor (428)	fair (183)	excellent	good (158)	fair
Knoxville	poor (554)	fair (255)	good (158)	excellent	poor

Note that the matrix is now nearly diagonally symmetric, except that Nashville still rates Memphis as "poor" because it is the furthest. Homunq (talk) 16:06, 6 March 2012 (UTC)

(1) I believe each voter's absolute scale should be calibrated independently by their own criteria. I believe we would be making an error by assuming each voter uses the same criteria which is conceived arbitrarily - i.e. "in miles, nice round intervals." Each voter's criteria is the proximity to their own city, which affects the calibration of their scale. The meaning of "absolute" in my understanding of Balinski's model is: (i) All voters are aware of the same range of possible alternatives (from their own vantage point), including those not in the given election (ii) The voter calibrates their own absolute scale considering the total range of possibilities, not just the alternatives in the election, so that a voter's ratings do not change when the alternatives available change. I believe the confusion may arise from mistaking the "common language" shared by voters (e.g. "Excellent","Poor") with each voter's "absolute" scale. Yes indeed the same language is known to all - i.e. that "Excellent" is at the to top the scale, and "Poor" at the bottom. But because voter's have different criteria, a "poor" rating is calibrated through the prism of their own criteria. If we were to assume that voters took into account other voters criteria, or a global criteria (if one exists), then it would seem to me in the example above the Memphis's perspective would calibrate the scale for everyone else, since Memphis voter's being on the edge of the state see the maximum differential between city distances, and so for all voters to be able to also express the same scale with a common language would need to adopt the Memphis absolute scale, or a scale considering the maximum distance from any two points on the state borderline. The problem with this direction, however, there may be no inter-comparability between two voters' criteria (e.g. access to drinking water VS hunting homeless people)

(2) A second problem I object to is the contradiction of proposing an "absolute" scale on the one hand, and then violating it by giving the least preferred alternative the lowest rating regardless of its "absolute" scale, presumably for the purpose of distinguishing it from the others. Why then would Memphis voters not be compelled to distinguish Chattanooga from Knoxville, for the same reason, which leads us back to ratings based on ranked preferences? Not to mention, this interpretation clearly violates the claim of IIA passage, a claim vehemently defended, that I do not agree with. (Please enter replies in new section "IIA Compatibility of Tennessee Example") Filingpro (talk) 11:20, 18 March 2016 (UTC)

I used normalized proportional values due to the fact that a voter that doesn't "know" a really long distance (because all cities are relatively nearby) would consider a distance of 200 miles as worse than a voter that "knows" long distances would.

But I'm OK with your version. Nevertheless, I think with your case the strategic considerations doesn't have to be made in detail, I only added the computation process since the "honest" case didn't had a tie. So, I propose to change the strategic considerations case to just one paragraph without the computation details (as in the version before my change ;) ) -- Arno Nymus, 2012-03-06 18:40 — Preceding unsigned comment added by 134.102.209.204 (talk)

Sounds good. Make it so. Thanks for starting this process off, by the way; I've added you on Google+. Homunq (talk) 17:49, 6 March 2012 (UTC)

You're welcome. But I have to add, that I don't have a Google+-account. -- Arno Nymus, 2012-03-07 21:00 — Preceding unsigned comment added by 77.23.79.151 (talk)

 Done I added a DONE-sign since this issue is resolved. Thx to Homunq. -- Arno Nymus (talk) 17:11, 15 March 2012 (UTC)

Reopened as per issues above (please enter replies in new section "IIA Compatibility of Tennessee Example") Filingpro (talk) 11:20, 18 March 2016 (UTC)