Loading ...
Sorry, an error occurred while loading the content.
 

Re: question and also introducing myself

Expand Messages
  • Bruce R. Gilson
    ... And that requires one other point being made, which he didn t state, but which is the major difficulty of Condorcet methods. A Condorcet method -- any
    Message 1 of 7 , Aug 1, 2008
      --- In RangeVoting@yahoogroups.com, Dave Ketchum <davek@...> wrote:
      >
      > On Thu, 31 Jul 2008 16:42:47 -0000 Joyce McCloy wrote:
      > > Hi. I'm Joyce McCloy, and mostly I have been visiting this group but
      > > hadn't joined. My primary activism efforts are towards verified voting
      > > issues in my home state of North Carolina. I have joined this group so
      > > that I may post questions to your experts.
      >
      > We AIN'T all experts, but you have demonstrated some expertise with
      > your questioning - enough to deserve thoughtful responses.
      >
      > I am a Condorcet backer [...]

      And that requires one other point being made, which he didn't state, but
      which is the major difficulty of Condorcet methods.

      A Condorcet method -- any Condorcet method -- tries to find the candidate
      who would beat all the others in a head-to-head matchup. If there _is_
      such a winner, fine and dandy. But often, there _is_ no such candidate.
      Even then, it can find the candidate who comes nearest to beating all the
      others. But occasionally it reaches a point where A could beat B (because
      C's backers favor A over B), B could beat C (because A's backers favor B
      over C), and C could beat A (because B's backers favor C over A). When
      this happens, we have a "Condorcet cycle." Because there is no
      _obvious_ winner in this case, _all_ Condorcet methods have to define a
      procedure for breaking the cycle, and these procedures are often so
      involved that an ordinary person cannot understand them. (Joyce, have
      someone describe the "Schulze method" to you, for example. If you can
      understand it, more power to you!)

      It is because of these cycles that can _only_ occur in Condorcet methods
      that I have decided that Condorcet methods are impracticable.
    • Juho
      ... In such situations some Condorcet methods put more emphasis on finding the best winner, some put more emphasis on minimizing the strategic threats. ... I
      Message 2 of 7 , Aug 1, 2008
        On Aug 1, 2008, at 18:19 , Bruce R. Gilson wrote:

        > --- In RangeVoting@yahoogroups.com, Dave Ketchum <davek@...> wrote:
        >>
        >> On Thu, 31 Jul 2008 16:42:47 -0000 Joyce McCloy wrote:
        >>> Hi. I'm Joyce McCloy, and mostly I have been visiting this group but
        >>> hadn't joined. My primary activism efforts are towards verified
        >>> voting
        >>> issues in my home state of North Carolina. I have joined this
        >>> group so
        >>> that I may post questions to your experts.
        >>
        >> We AIN'T all experts, but you have demonstrated some expertise with
        >> your questioning - enough to deserve thoughtful responses.
        >>
        >> I am a Condorcet backer [...]
        >
        > And that requires one other point being made, which he didn't
        > state, but
        > which is the major difficulty of Condorcet methods.
        >
        > A Condorcet method -- any Condorcet method -- tries to find the
        > candidate
        > who would beat all the others in a head-to-head matchup. If there _is_
        > such a winner, fine and dandy. But often, there _is_ no such
        > candidate.
        > Even then, it can find the candidate who comes nearest to beating
        > all the
        > others.

        In such situations some Condorcet methods put more emphasis on
        finding the best winner, some put more emphasis on minimizing the
        strategic threats.

        > But occasionally it reaches a point where A could beat B (because
        > C's backers favor A over B), B could beat C (because A's backers
        > favor B
        > over C), and C could beat A (because B's backers favor C over A). When
        > this happens, we have a "Condorcet cycle."

        I wouldn't call the cycles "Condorcet cycles" since they are not
        Condorcet specific. They are a general property of group opinions,
        and are not generated by any election method. Some Condorcet methods
        handle them explicitly, and when the Condorcet criterion does not
        directly determine the winner then there must be at least one cycle
        in the group opinions, but that should not be counted against them.

        > Because there is no
        > _obvious_ winner in this case, _all_ Condorcet methods have to
        > define a
        > procedure for breaking the cycle,

        Some Condorcet methods have a specific (explicit) cycle breaking
        procedure, some don't.

        > and these procedures are often so
        > involved that an ordinary person cannot understand them.

        Some Condorcet methods are so much "tuned" that they are relatively
        complex to understand to regular voters. The underlying Condorcet
        principle should however be simple enough.

        There are also Condorcet methods that are quite easy to understand,
        offer a natural explanation to the results, and that do not have
        specific cycle breaking part. The most obvious such method is one
        that elects the candidate that would need least additional votes (if
        any) to beat all others. That method is usually referred to as "minmax
        (margins)".

        > (Joyce, have
        > someone describe the "Schulze method" to you, for example. If you can
        > understand it, more power to you!)
        >
        > It is because of these cycles that can _only_ occur in Condorcet
        > methods
        > that I have decided that Condorcet methods are impracticable.

        The cycles occur in the group opinions and are thus present in all
        election methods. Different methods just treat them in different ways
        (explicitly or implicitly).

        The point is that even if all the voters (in any method) are expected
        to have linear preferences like A>B>C>D>E, when one sums up the
        opinions of a group of voters one may find out that the group
        (majority of the members) prefers A over B, B over C and C over A. No
        voting method can fix that problem (since it is part of the laws of
        nature). What this can mean in practice is that after one of the
        candidates (let's say there are only three of them and they form a
        cycle) has been elected there can be strategic voters who think their
        favourite candidate shouldn't have even participated the election
        since that would have improved the outcome (the second best candidate
        would have won instead of the worst), or that they should have voted
        dishonestly to improve the outcome.

        Juho


        >
        >
        > ------------------------------------
        >
        > Yahoo! Groups Links
        >
        >
        >



        ___________________________________________________________
        Copy addresses and emails from any email account to Yahoo! Mail - quick, easy and free. http://uk.docs.yahoo.com/trueswitch2.html
      • Dave Ketchum
        Agreed that cycles are a major Condorcet debate item, that they do complicate life a bit, and that I did not mention them in my last post. HOWEVER, they are
        Message 3 of 7 , Aug 1, 2008
          Agreed that cycles are a major Condorcet debate item, that they do
          complicate life a bit, and that I did not mention them in my last post.

          HOWEVER, they are not exactly as Bruce describes here.

          They are three or more candidates in near ties, supported by multiple
          groups of conflicting voters.

          btw - there either is a single winner or there is a cycle (can manage
          a true tie - hard with real voters).

          This is not unique to Condorcet.
          IRV can read exactly the same ballots and see no problem, for at
          any moment it is concerned with discarding only what it sees as the
          single weakest candidate.
          Plurality can see candidates in near ties and do runoffs without
          caring about details.

          Looking closer at a simple cycle of A>B, B>C, and C>A, and 3 voters:
          vote counts
          A>B A>B B>C A>C
          B>C B>A B>C C>A
          C>A A>B C>B C>A
          totals: 2A>B+B>A 2B>C+C>B 2C>A+A>C

          Above is a perfect tie among equal sized groups. Among the possible
          complications are B>A, C>B, and A>C.

          I claim these will not be too common. Note that two groups cannot
          make a cycle - it takes three or more with near equal strength - think
          of how often you get three near-equal candidates in Plurality - with
          the conflicting goals.

          I also claim the average voter does not have to understand all the
          details:
          Takes plenty of debate to decide on method.
          Voter needs to know that this has been done, and see some
          examples of how it works.

          DWK

          On Fri, 01 Aug 2008 15:19:41 -0000 Bruce R. Gilson wrote:
          > --- In RangeVoting@yahoogroups.com, Dave Ketchum <davek@...> wrote:
          >
          >>On Thu, 31 Jul 2008 16:42:47 -0000 Joyce McCloy wrote:
          >>
          >>>Hi. I'm Joyce McCloy, and mostly I have been visiting this group but
          >>>hadn't joined. My primary activism efforts are towards verified voting
          >>>issues in my home state of North Carolina. I have joined this group so
          >>>that I may post questions to your experts.
          >>
          >>We AIN'T all experts, but you have demonstrated some expertise with
          >>your questioning - enough to deserve thoughtful responses.
          >>
          >>I am a Condorcet backer [...]
          >
          >
          > And that requires one other point being made, which he didn't state, but
          > which is the major difficulty of Condorcet methods.
          >
          > A Condorcet method -- any Condorcet method -- tries to find the candidate
          > who would beat all the others in a head-to-head matchup. If there _is_
          > such a winner, fine and dandy. But often, there _is_ no such candidate.
          > Even then, it can find the candidate who comes nearest to beating all the
          > others. But occasionally it reaches a point where A could beat B (because
          > C's backers favor A over B), B could beat C (because A's backers favor B
          > over C), and C could beat A (because B's backers favor C over A). When
          > this happens, we have a "Condorcet cycle." Because there is no
          > _obvious_ winner in this case, _all_ Condorcet methods have to define a
          > procedure for breaking the cycle, and these procedures are often so
          > involved that an ordinary person cannot understand them. (Joyce, have
          > someone describe the "Schulze method" to you, for example. If you can
          > understand it, more power to you!)
          >
          > It is because of these cycles that can _only_ occur in Condorcet methods
          > that I have decided that Condorcet methods are impracticable.
          --
          davek@... people.clarityconnect.com/webpages3/davek
          Dave Ketchum 108 Halstead Ave, Owego, NY 13827-1708 607-687-5026
          Do to no one what you would not want done to you.
          If you want peace, work for justice.
        • raphfrk
          ... Another option would be to do the count in a 2 stage process. In the first step, the condorcet winner of the candidates on the ballot is determined. Once
          Message 4 of 7 , Aug 2, 2008
            --- In RangeVoting@yahoogroups.com, Dave Ketchum <davek@...> wrote:
            >
            > Write-ins can complicate life a bit. Not intolerably. Provided
            > proper preparation is done, when there is need to accommodate an extra
            > candidate, the matrix can be copied into one with the needed extra
            > column and row, filled in with exactly the same data as if known at
            > the beginning of counting.

            Another option would be to do the count in a 2 stage process. In the
            first step, the condorcet winner of the candidates on the ballot is
            determined.

            Once that has been finalised, each polling station counts each
            write-in vs the provisional winner.

            In most cases, none of the write-ins would defeat the provisional
            winner. In the unlikely event that one does, then just declare the
            write in the winner (which might be a little unfair).

            If there is more than 1 write in who beats the provisional winner (or
            for the sake of fairness), a new matrix would be created containing
            all the write ins who defeated the provisional winner and also any
            candidates in the Smith set. Each polling station would fill in those
            totals.
          • Dave Ketchum
            Better, treat all write-ins as a single candidate. That candidate will usually be proved a loser with little pain or effort. If not, punt to smarter code -
            Message 5 of 7 , Aug 2, 2008
              Better, treat all write-ins as a single candidate. That candidate
              will usually be proved a loser with little pain or effort. If not,
              punt to smarter code - there could be many write-in names.

              DWK

              ON Sat, 02 Aug 2008 20:19:26 -0000 raphfrk wrote:
              > --- In RangeVoting@yahoogroups.com, Dave Ketchum <davek@...> wrote:
              >
              >> Write-ins can complicate life a bit. Not intolerably. Provided
              >>proper preparation is done, when there is need to accommodate an extra
              >>candidate, the matrix can be copied into one with the needed extra
              >>column and row, filled in with exactly the same data as if known at
              >>the beginning of counting.
              >
              >
              > Another option would be to do the count in a 2 stage process. In the
              > first step, the condorcet winner of the candidates on the ballot is
              > determined.
              >
              > Once that has been finalised, each polling station counts each
              > write-in vs the provisional winner.
              >
              > In most cases, none of the write-ins would defeat the provisional
              > winner. In the unlikely event that one does, then just declare the
              > write in the winner (which might be a little unfair).
              >
              > If there is more than 1 write in who beats the provisional winner (or
              > for the sake of fairness), a new matrix would be created containing
              > all the write ins who defeated the provisional winner and also any
              > candidates in the Smith set. Each polling station would fill in those
              > totals.
              --
              davek@... people.clarityconnect.com/webpages3/davek
              Dave Ketchum 108 Halstead Ave, Owego, NY 13827-1708 607-687-5026
              Do to no one what you would not want done to you.
              If you want peace, work for justice.
            Your message has been successfully submitted and would be delivered to recipients shortly.