[Koha] Koha license upgrade voting method

[Thomas Dukleth]

I am pleased that Clay Shentrup, an especially well informed very strong
advocate for score voting (range voting), has contributed to the
discussion.  I am grateful for his links to informative analyses and his
reference to the William Poundstone book "Gaming the Vote".  Score voting
as a name seems to convey the meaning of the method better, although,
range voting is the most popular name for the method.

I have read some of William Poundstone's books in past decades with great
interest.  His book on voting methods is an especially up to date popular
introduction.  Gaming the vote : why elections aren't fair (and what we
can do about it) / William Poundstone. - New York : Hill and Wang, 2008. 
The  adoption of voting methods using software to analyse preference votes
in online communities such as Debian is mentioned.  The most significant
criticism I have of what I have read from the book is that Poundstone
sometimes accepts a little too much from some voting theory experts at
face value without more careful analysis.  [Clay is described in the book
as a fanatical advocate for score voting but good ideas need somewhat
fanatical supporters to ensure that they have due attention.  Without
strong support from some people, good ideas are often overlooked in favour
of what is already popular or well financed.]

While score voting is commonplace in the world examining preferences, its
explicit proposal as a voting method for elections may be dated from a
paper by Warren Smith introducing the idea.  Range Voting / Warren D
Smith. - December 2000 -
http://www.math.temple.edu/~wds/homepage/rangevote.pdf .

Score voting may be interpreted as a superset of multiple voting methods
including the highly regarded Condorcet method,
http://www.scorevoting.net/CondDQ.html .  Score voting more consistently
obtains a winner conforming to the Condorcet principle than the classic
Condorcet method does itself.  The Condorcet principle sums the winners of
the preferences for each candidate compared to each other candidate in a
series of candidate pairs.

In another message, I will address the important issues of voting strategy
with score voting.


Remainder of reply inline:

On Sat, January 8, 2011 04:38, Clay Shentrup wrote:


1.  CONSTRICTING RANGE RELATIVE TO NUMBER OF VOTERS.

> FairVote's argument against Score Voting (aka Range Voting) is
> mathematically
> flawed.

I had not remembered anything about FairVote and had merely observed the
link from the range voting Wikipedia article.  The deficiency in the
FairVote criticism of range voting to which I linked was that the scenario
they give uses a range which is unduly high relative to the number of
voters.

Their scenario has 100 voters casting ballots with a range of 0 to 99. 
The relatively large range allows merely two voters using a reasonable
voting strategy of casting their greatest preference strongly at the
maximum range of 99 to outvote all other voters expressing their greatest
preference weakly near the minimum of the range at 1.

The construction of the case is so unreasonable that I expected I was
missing something about how score votes are meant to be counted.

Score voting is concerned with voting within a restricted range.  Without
any range, whichever voter could record the highest imaginable number on a
ballot would decide an election.

A reasonably low range relative to the number of voters along with voters
who use some half intelligent sense of voting strategy avoids the
unreasonable case above.  [In a later message, I will propose a means for
resolving some issues of voting strategy.]


2.  FAIRVOTE: THE CENTER FOR VOTING AND DEMOCRACY.

> FairVote began as "Citizens for Proportional Representation", and
> I
> believe their real interest in implementing Single Transferable Vote in
> the
> USA, like e.g. Australia uses in their Senate. Because IRV ("preferential
> vote") is the single-winner case of STV, FairVote pushes it as a useful
> "stepping stone". I do not believe they particularly care about the merits
> of
> IRV, and in fact they have made numerous blatantly false and misleading
> statements on voting for many years.

Some statements from FairVote founder, Robert Richie, seem to show that he
would advocate whatever non-plurality (non-first past the post) voting
method  would be most likely to obtain sufficient political support to be
implemented for government elections.  Any real preference by FairVote may
be for a voting method which is useful as first step to proportional
representation, the long term objective of FairVote.  Instant runoff
voting can be transformed into the single transferable ballot for
proportional representation.

Poundstone reports that there has been sufficient support to institute
instant runoff voting in some jurisdictions because of well financed
support from FairVote.  Less well financed supporters of alternatives
other than instant runoff voting appear divided because all voting systems
are subject to some significant criticism.


3.  INSTANT RUNOFF VOTING.

>
> Warren D. Smith, a Princeton math Ph.D. who has studied voting theory for
> over a decade, and sort of served as the "protagonist" of the book
> _Gaming the Vote_ responds specifically to several of the
>
> FairVote statements here:
> http://ScoreVoting.net/Irvtalk.html

My innocent and ignorant link to the FairVote criticism of range voting
within their comparison of instant runoff voting (IRV) to other voting
systems had not been meant as a suggestion that instant runoff voting be
considered for the Koha community.  I had considered making a comment
against IRV along with my link to the FairVote criticism of range voting
but I decided against that in case FairVote might have some adequate
modification of IRV.

I have not identified any adequate modification of IRV.  IRV works by
always eliminating some preferences.  [Other preference voting systems
attempt to compare all preferences without eliminating any.]  Eliminating
preferences is a poor basis for a goal of maximising voter preferences and
leads to an unacceptably high likelihood of a result which people would
find unwanted.


4.  SCORE VOTING (RANGE VOTING).

>
> Here is a response to the specific (and actually quite absurd) criticism
> that
> Score Voting and Approval Voting violate "majority rule", particularly as
> compared to IRV.
>
> http://www.electology.org/majority
>
> The "right" metric of voting method performance (and you can basically
> prove it
> mathematically, as counter-intuitive as that may seem) is "average voter
> satisfaction" which can be expressed using ametric called Bayesian regret.
> Look
> at this graph of Bayesian regret values and notice how vastly much better
> Score
> and Approval Voting perform. IRV is actually almost bad as Plurality/FPtP.
> http://scorevoting.net/BayRegsFig.html

The calculation of Bayesian regret in the graph cited from "Gaming the
Vote" is a selection from Warren Smith's computer simulation of 144 voting
scenarios with different parameters in which regret is presumably
triggered by outcomes disfavouring and sometimes subverting preferences of
individual voters in the simulation.

The election simulation software used is the Infinitely Extendible Voting
Simulator (IEVS).  The best description which I found for IEVS is in a
grant proposal, http://www.rangevoting.org/PewIEVS.pdf .  The code is
available at http://www.rangevoting.org/IEVS/IEVS.c .


Thomas Dukleth
Agogme
109 E 9th Street, 3D
New York, NY  10003
USA
http://www.agogme.com
+1 212-674-3783