[EM] Majority-Judgement using adjectives versus alphabetical scales versus numerical ranges.

[Kristofer Munsterhjelm]

On 12/08/2012 06:19 AM, ⸘Ŭalabio‽ wrote:
> 0.
>
> ¿What is with all of the people top-posting?:
>
> http://web.archive.org/web/20080113211450/http://www.greenend.org.uk/rjk/2000/06/14/quoting.html

They do it because modern mail software makes that the default. I think 
top-posting is a bit distracting, too, but the fault is Microsoft and 
Google's, not the posters'.

Why Google and Microsoft makes their software default to top-posting, I 
don't know.

> 1.
> I received some strange suggestions for ranges:
>
> *	Negative -10 to positive +10
> *	Negative -5 to positive +5
> *	Negative -100 to positive +100
> *	1 to 5
> *	1 to 10
> *	1 to 100
>
> If we shall use 2-digit numbers like 10, we might as well use all
> 2-digit numbers up to 99. Otherwise, one should just go to 9 and stop.
> If we have 100, we might as well go to 999 or stop at 99.
>
> Stopping at 5 makes no sense. If we use 1-digit numbers, we should
> use  all 1-digit numbers.
>
> The ranges starting with 1 have 3 terrible things wrong them:
>
> 0. If we use positive real number, ¡I want to give that fascist a 0!
> ¡Giving that jerk 1 point is 1 point too many!
> 1. If we use a certain sized digit-range, we should use all of the
> range.
> 2. With the ratings starting at 1, the voters might get confused and
>rank instead of rate with the best candidates rated at 1 and the worst
>at n which is backwards.
>
> This is my preferred range:
>
> Negative -99 to positive +99

The ponies already objected to your preferred range, and I think their 
objection has validity. If they find it too hard to find the right 
rating between -99 and +99, then they'll consider the method bad however 
you put it. Again, RBJ has voiced the same point here on the list: 
"Range asks for too much, Approval asks for too little".

You argued in turn that this wasn't so important, and that they could 
just not use the granularity if they didn't need it, but that didn't 
convince them either.

Thus I suggest that (even if it is your preferred method and range), 
that isn't going to work for the ponies. The objective quality of the 
objection doesn't matter - it keeps them from using the method because 
they don't like using it :-)

And hence I suggested MJ (which works better with few grades, not so few 
as to make it Approval and not so many as to overwhelm the ponies), or 
Schulze/[insert Condorcet method of choice] if they want ranking instead 
of rating.

> 2.
>
> Some believe that we should use adjectives instead of letters
> because  the voters understand that better. This may be true in France, but is
> _“*NOT*”_ true in the United States Of America:
>
> In the United States Of America, over the course of a dozen years,
> teachers grade students on the scale of from A+ to F- over 10 thousand
> times. Americans have had this scale pounded into their heads so many
> times that they understand it all too well.
>
> I imagine that, for Americans, using adjectives should be harder
> than  using the alphabetical scale of A+ though F-.

There's another point to using letter grades or adjectives instead of 
numbers. Doing so shows that it doesn't really matter what kind of 
points scale the grades are calibrated to, as long as a majority shares 
the same calibration. That is in turn linked to what I said about 
monotone nonlinear transformations.

And MJ itself does tiebreaking by providing + and - to the actual grade 
returned. For instance, in the 2007 Orsay experiment, Bayrou got Good+ 
while Sarkozy got Good-. So I don't think the grade inputs should be 
given with + and -, because the method appends a + and - to the result 
as part of the tiebreak process.