[EM] The election methods trade-off paradox/impossibility theorems paradox.

[fdpk69p6uq at snkmail.com]

On Fri, Jun 23, 2017 at 1:02 AM, robert bristow-johnson wrote:

> but voters can skew that information quantitatively and, if they want to
> be tactical, insincerely.
>

...and even when *everyone* votes tactically, rated systems still
outperform ranked systems.

http://electology.org/sites/default/files/comparing_voting_m
ethods_simplicity_group_satisfaction.png

https://electology.github.io/vse-sim/vse.html

"Tactical Score Voting may be, in practice, more likely to elect a
Condorcet winner (when one exists) than real Condorcet methods."
https://electology.org/tactical-voting-basics

if i really, really, really like Candidate A over Candidate B and you only
> sorta like B over A by just a little bit, it should not matter to what
> disparate degree we like our candidates.
>

Yes, it absolutely should.  That's the fundamental flaw in ranked ballots:
preference rankings can't be meaningfully compared between individuals. If
there are 2 voters and 2 candidates, and the ballots are:

1: A > B
2: B > A

you can't tell who the winner should be, but if their true feelings are

1. loves A and hates B
2. loves B and likes A

the obviously correct winner is A.

The goal of a voting system is to choose the candidate who best represents
the population / whose opinions are closest to the centroid of the
population's opinions / who, winning, would maximize the happiness of the
population.  Ranked ballots destroy information that's needed to find this
winner.


On Fri, Jun 23, 2017 at 8:35 AM, Brian Olson wrote:

> I don't favor raw Score summation. It's strategy prone. For choices where
> my honest vote might be [1.0, 0.8, 0.6, 0.4, 0.2, 0.0] I should probably
> vote strategically [1.0, 1.0, 1.0, 0.0, 0.0, 0.0].
>

... in which case it devolves into Approval Voting, which... is still
pretty good.

Anyway, this is only your optimal strategy if you know exactly what
everyone else is voting.  In reality, polls have uncertainty (e.g. November
2016), and ratings ballots allow you to hedge your bets based on the
likelihood of different outcomes.

And if you don't like that and the varying vote power depending on how you
> vote: I have a system for you!
> "Instant Runoff Normalized Ratings" (IRNR)
>

Do you think this has a benefit over SRV/STAR?


On Fri, Jun 23, 2017 at 12:09 PM, robert bristow-johnson wrote:

> No, we should not make the voters cook up that information.  all we should
> ask the voters is "whom do you prefer A or B?"  and "if you can't get your
> favorite, whom is your next preference?"
>

Utility information is the information that's naturally in voters' brains.
Ranking ballots force them to "quantize" it into discrete steps which
distort the underlying reality. If a voter's true feelings are:

A: Love
B: Like
C: Hate
D: Hate
E: Hate
F: Hate
G: Hate

a ranking ballot forces them to vote something like

A: 1
B: 2
C: 3
D: 4
E: 5
F: 6
G: 7

amplifying small preferences into large ones and making it look like they
approve of C 5/7ths as much as they approve of A.  When you combine this
across many voters, C is going to get much more power than they actually
deserve.

"The majority judgement experiment proves that the model on which the
theory of social choice and voting is based is simply not true: voters do
not have preference lists of candidates in their minds. Moreover, forcing
voters to establish preference lists only leads to inconsistencies,
impossibilities and incompatibilities." https://hal.archives-ouvertes.
fr/hal-00243076

"Thus, the discrimination forced by rankings lessened the validity of
rankings among participants who did not freely differentiate between
values. In other words, our findings indicate that ratings may have more
predictive validity than rankings, perhaps because the latter force
participants to sometimes make unimportant and/or inconsequential (and
hence invalid) distinctions between similarly regarded values. ... value
ranking forces people to make distinctions that they would not otherwise
make" http://dx.doi.org/10.1207/s15324834basp1802_4

"cardinal (utility-based) preferences are embedded into the space of
ordinal preferences. This often gives rise to a distortion in the
preferences, and hence in the social welfare of the outcome."
http://procaccia.info/papers/distortion.cia06.pdf

no reason to use that over Condorcet (with a simple method to deal with
> cycles, ranked-pairs is still a lot easier to explain than Schulze and they
> pick the same winner if there are 3 in the Smith set, so let's use the
> simpler method)
>

Condorcet systems are too complicated for real-life use in governmental
elections, and much simpler cardinal systems produce the same or better
outcomes in practice.

(Though the goal should be to find the most-approved Utilitarian Winner,
not the most-preferred Condorcet Winner, but they are the same in most
cases. http://www.sciencedirect.com/science/article/pii/S0304406815000518)


On Fri, Jun 23, 2017 at 2:28 PM, Richard Lung wrote:

>
> I also have found it hard to believe that Score voting and Approval voting
> are taken seriously.
>


and I find it hard to believe that people are still seriously arguing about
which ranked, majoritarian voting methods meet which mathematical criteria,
when rated utilitarian voting methods exist, and are accepted to work just
fine in every other context in which they are used.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20170623/a6ba6114/attachment.htm>