Re-Proposal -- Change constitution to adopt Smith/Condorcet vote tallying
On Sat, Dec 16, 2000 at 01:36:18AM +1000, Anthony Towns wrote:
> So, uh, what happened to:
> > Personally, I worry about any kind of wholesale change in the language
> > of the constitution. Yeah, if you change major chunks of the document
> > then current ambiguities would go away. But how do we know whether we're
> > introducing new ones?
I looked at the problems we had, looked where minimal language
fixes would take us, and decided to agree with you on the advantages
of Smith/Condorcet. What I'm trying for, now, is a system which is
equivalent to the current system for cases where interpretation of the
current system is not ambiguous.
The "minimal language fixes" which I could come up with, which addressed
all the ambiguities in the tallying system, without losing the advantages
of the tallying system, wound up being more wordy and complex than the
rewrite. [I didn't want to lose the ability to deal with multi-issue
ballots. I think that forcing an issue to be split across multiple
ballots tends to force solutions to focus more on the way voting is
carried out and less on the issue itself.]
However, as you can see, I've already withdrawn and reproposed the
Smith/Condorcet change a number of times, to fix various non-obvious
problems. I'm doing it again in this message. So: you have a point,
introducing a change of this magnitude is less than ideal.
> What you describe looks something like one of the Minmax variants or
> the Tideman method described at:
> It doesn't use the Smith criterion directly. It's not particularly obvious
> that it even satisfies the Smith criterion, afaict.
First: I added a step to satisfy the smith criterion, earlier this
morning. [It now satisfies smith after the first time through the
Second: Minmax "winning-votes" is based on number of votes against,
and doesn't consider the number of votes for an option, but "votes
for" is used to organize the winning criteria table in my proposal.
Minmax "margins" is based on the difference between votes for and votes
against, but I never examine the difference between the votes for and
the votes against in my proposal. Why would you think that my proposal
is equivalent to either of these.
Third: Tideman works from the top down, not the bottom up. The method
I proposed works from the bottom up. I don't see how you could get an
equivalence between my proposal and Tideman, either.
On the other hand: I think Tideman is better than using casting vote for
resolving ties, so I'm withdrawing my proposal from earlier this morning
and re-issuing it with a mechanism taken from Tideman for certain kinds
So: I'm withdrawing my proposal from earlier this morning, and
proposing the following in its place:
--- /usr/share/doc/debian/constitution.txt Tue Sep 14 18:00:00 1999
+++ /tmp/constitution.txt Fri Dec 15 11:21:55 2000
@@ -197,9 +197,9 @@
nominated themselves and have not yet withdrawn, plus None Of The
Above. If None Of The Above wins the election then the election
procedure is repeated, many times if necessary.
- 7. The decision will be made using Concorde Vote Counting. The quorum
- is the same as for a General Resolution (s.4.2) and the default
- option is None Of The Above.
+ 7. The decision will be made using Smith/Condorcet Vote Counting. The
+ quorum is the same as for a General Resolution (s.4.2) and
+ the default option is None Of The Above.
8. The Project Leader serves for one year from their election.
@@ -259,7 +259,7 @@
including themselves; there is no None Of The Above option. The
vote finishes when all the members have voted or when the outcome
is no longer in doubt. The result is determined according to
- Concorde Vote Counting.
+ Smith/Condorcet Vote Counting.
8. The Chairman can stand in for the Leader, together with the
As detailed in s.7.1(2), the Chairman of the Technical Committee
@@ -517,9 +517,9 @@
elsewhere. If the voting period can end if the outcome is no
longer in doubt, the possibility that voters may change their
votes is not considered.
- 5. The votes are counted according to the Concorde Vote Counting. If
- a quorum is required then the default option is Further
+ 5. The votes are counted according to the Smith/Condorcet Vote
+ Counting. If a quorum is required then the default option is
+ Further Discussion.
6. In cases of doubt the Project Secretary shall decide on matters of
procedure (for example, whether particular amendments should be
considered independent or not).
@@ -544,48 +544,59 @@
otherwise dealt with for 4 weeks then it is considered to have been
- A.6. Concorde Vote Counting
+ A.6. Smith/Condorcet Vote Counting
1. This is used to determine the winner amongst a list of options.
Each ballot paper gives a ranking of the voter's preferred
options. (The ranking need not be complete.)
- 2. Option A is said to Dominate option B if strictly more ballots
- prefer A to B than prefer B to A.
- 3. All options which are Dominated by at least one other option are
- discarded, and references to them in ballot papers will be
- 4. If there is any option which Dominates all others then that is the
- 5. If there is now more than one option remaining Single
- Transferrable Vote will be applied to choose amongst those
- + The number of first preferences for each option is counted,
- and if any option has more than half it is the winner.
- + Otherwise the option with the lowest number of first
- preferences is eliminated and its votes redistributed
- according to the second preferences.
- + This elimination procedure is repeated, moving down ballot
- papers to 2nd, 3rd, 4th, etc. preferences as required, until
- one option gets more than half of the `first' preferences.
- 6. In the case of ties the elector with a casting vote will decide.
- The casting vote does not count as a normal vote; however that
- elector will usually also get a normal vote.
- 7. If a supermajority is required the number of Yes votes in the
- final ballot is reduced by an appropriate factor. Strictly
- speaking, for a supermajority of F:A, the number of ballots which
- prefer Yes to X (when considering whether Yes Dominates X or X
- Dominates Yes) or the number of ballots whose first (remaining)
- preference is Yes (when doing STV comparisons for winner and
- elimination purposes) is multiplied by a factor A/F before the
- comparison is done. This means that a 2:1 vote, for example, means
- twice as many people voted for as against; abstentions are not
- 8. If a quorum is required, there must be at least that many votes
- which prefer the winning option to the default option. If there
- are not then the default option wins after all. For votes
- requiring a supermajority, the actual number of Yes votes is used
- when checking whether the quorum has been reached.
+ 2. A square "initial totals table" is constructed, recording the
+ ballot totals: the number in the table at row j, column k indicates
+ the number of ballots which prefer ballot option j to ballot
+ option k.
+ 3. A square "adjusted totals table" is constructed such that all
+ entries in row j from the initial totals table have been multiplied
+ by m and divided by n where option j has an n:m majority
+ requirement. For options with no such majority requirement,
+ the row in the adjusted totals table is the same as the row from
+ the initial totals table.
+ 4. A four column "winning criteria table" is constructed with one row
+ for each cell in the adjusted totals table: the "for" column
+ holds option j; the "against" column holds option k; the "votes
+ in favor" column holds the value from the adjusted totals table
+ for row j, column k; the "votes against" column holds the value
+ from the adjusted totals table for row k, column j.
+ 5. The winning criteria table is sorted in descending order based on
+ the "votes in favor" column (and, where values in the "votes in
+ favor" column are equal, in ascending order based on the "votes
+ against" column).
+ 6. Eliminate all rows from the winning criteria table where "votes
+ in favor" are less than the corresponding "votes against" column.
+ 7. If there is a single option in "for" column of the winning criteria
+ table which does not appear in "against" column, that option is
+ the chosen option.
+ 8. If there are multiple options in the "for" column of the chosen
+ criteria table which do not appear in "against" column, the
+ option(s) which correspond to rows in the winning criteria table
+ where "votes in favor" and "votes against" equivalent to those
+ of the first row are candidates for being chosen. If there's
+ more than one candidate for being chosen, the person with a
+ casting vote picks the chosen option from among the options listed
+ in the "for" column of the candidate rows. If there's only one
+ candidate, the option listed in the "for" column of the candidate
+ row is the chosen option.
+ 9. Until there is a chosen option, the last row of the winning
+ criteria table is examined and (permanently) removed from the
+ winning criteria table. At the same time, any adjacent row(s) are
+ removed, if they are identical to what the removed row has in the
+ "votes in favor" and "votes against" columns. Exception: if all
+ rows of the winning criteria would be eliminated, the person with
+ a casting vote picks the chosen option from those which remained
+ in the "for" column.
+ 10. Once the chosen option is picked, the ballots are re-checked:
+ if the number of ballots which mention the chosen option is less
+ than a quorum for the vote, the default option wins. Otherwise,
+ the chosen option wins.
When the Standard Resolution Procedure is to be used, the text which
refers to it must specify what is sufficient to have a draft
resolution proposed and/or sponsored, what the minimum discussion