2017 update to the SPI voting algorithm for Board elections [and 1 more messages]

Ian Jackson ijackson at chiark.greenend.org.uk
Mon Feb 27 18:09:21 UTC 2017


Dimitri John Ledkov writes ("2017 update to the SPI voting algorithm for Board elections"):
> The board recognizes the deficiencies of the current voting algorithm
> utilized for Board elections as being inappropriate for multi-seat
> elections. After careful consideration, we recognize and acknowledge
> the evidence presented and we support updating the voting algorithm.

Thanks.

> Given that the voting algorithm is established via a Board resolution
> and given the extensive analysis undertaken by Ian Jackson, we invite
> Ian to draft a resolution to replace the existing Condrocet-based
> voting algorithm with a more appropriate algorithm, taking into
> account any existing feedback from spi-general, in time for the March
> meeting (2017-MAR-13).

I therefore intend to draft a resolution to replace the Condorcet
board voting system with the Single Transferrable Vote (STV).

There are numerous variants of STV, which have minor differences.  The
differences only matter in edge cases, but they affect the detailed
calculations at each stage.  It is necessary to pick a variant.

Candidate variants which have been plausibly suggested are:

* The UK Electoral Reform Society's Rules for STV.  This is very
  well-established, but it has many features which are optimised for
  hand-counting with paper ballots.  These edge case interpretations
  are tiresome to program in a computer and sometimes a bit odd.

* Scottish STV, as used by local government elections in Scotland.
  This has several advantages:
   - There is a clear, readable and unambiguous specification
     set out in public UK legislation.
   - There is an implementation in OpenSTV, which seems reliable
     (and has been around for quite some time)
   - As an experiment, I reimplemented the algorithm myself and
     found it to be tractable

* Meek STV, as used by the Apache Software Foundation.  In my searches
  I didn't manage to find a stable and widely-used implementation.  I
  am also more comfortable with adopting a voting system choice from
  politically-oriented parts of civil society, or governments, than I
  am from software organisations.  Many software organisations have a
  poor record of handling governance questions.

I therefore intend to propose a resolution that we adopt Scottish STV.
Here's a draft.


WHEREAS

1. Condorcet is suitable for a single-winner voting system, but is
   seriously lacking in proportionality in for multi-winner voting
   systems, such as SPI's Board Elections.

2. The Single Transferrable Vote is a very widely-adopted proportional
   preferential voting system.  (And may be the only such system.)

3. The Scottish STV variant is clearly specified; we have an
   established and stable Free Software implementation of it; and it
   is straightforward to (re)implement.  Other STV variants appear to
   lack some of these good properties.

4. Ian Jackson has offered to help with the implementation of STV for
   SPI.

THEREFORE THE SPI BOARD RESOLVES

5. Future elections to the SPI Board will be counted according to the
   Scottish Single Transferrable Vote.  Scottish STV will also be
   used by SPI for any other multi-winner election.

6. Specifically, the algorithm to be used is that specified in
   Rules 45-52 of the Scottish Local Government Elections Order
   2007 (a UK Statutory Instrument):
      http://www.legislation.gov.uk/ssi/2007/42/schedule/1/part/III/crossheading/counting-of-votes/made

7. We expect that the practical implementation will be by means of
   OpenSTV, for example the openstv package in Debian.  However, any
   differences between the Rules in the Order, and the implementation
   in OpenSTV (or other software), are to be resolved in favour of the
   Rules.

8. The SPI Secretary is requested to liase with Ian Jackson, so that
   the necessary changes to SPI software and infrastructure can be
   identified and implemented.


Joshua D. Drake writes ("Re: 2017 update to the SPI voting algorithm for Board elections"):
> On 02/25/2017 08:46 AM, Dimitri John Ledkov wrote:
> > Given that the voting algorithm is established via a Board resolution
> > and given the extensive analysis undertaken by Ian Jackson, we invite
> > Ian to draft a resolution to replace the existing Condrocet-based
> > voting algorithm with a more appropriate algorithm, taking into
> > account any existing feedback from spi-general, in time for the March
> > meeting (2017-MAR-13).
> 
> This is great news. Ian, question... is there anything wrong with using 
> what OpenSource.org uses? (https://heliosvoting.org/)

That's a website.  There is no documentation explaining what their
system is.  The hamburger in the top right didn't do anything for me
(perhaps because I have JavaScript disabled).

I did a bit of web searching and this seems to be some kind of novel
ballot casting protocol.  I don't even see any discussion of what
voting system[1] they use or support.

Let's fix our voting system[1] first and then think about improving
our ballot casting protocol.

Thanks,
Ian.

[1] I mean, the algorithm used to convert a pile of ballot papers into
a result.  Not the voting protocol.


More information about the Spi-general mailing list