Skip to main content

View Diary: Gaming the vote - a book review and intro to voting theory (with poll) (58 comments)

Comment Preferences

  •  Instant runoff voting is the best (3+ / 0-)
    Recommended by:
    rilkas, neroden, BYw

    system that's currently used for voting anywhere.

    But I think range voting is better.

    •  right (3+ / 0-)
      Recommended by:
      Wee Mama, plf515, BYw

      Like the previous poster I hadn't heard of range voting.

      One thing I would like to see before supporting that system is some psychological studies of how people apply such metrics. It may be easier to apply numbers than to rank candidates, but easier doesn't necessarily mean better, nor does it necessarily entail greater accuracy.

      With numbers you could have

      Voter A: extremist temperament -- votes 100 for favorite candidate and 3, 2, 1 for all the others

      Voter B: moderate temperament -- votes 100 for favorite, but 70, 30, 20 for the others.

      Voter C: unenthusiastic temperatment -- votes 20, 10, 5, 1

      Voters A , B and C would be conflated on the absolute ranking scheme, but their votes are radically different on the range scheme.

      The question is whether the range scheme more accurately portrays voter preferences or merely portrays voters' temperaments in the number assignment task.

      In principle, the range scheme could permit a candidate whom many people gave a moderate number to, to beat a candidate who a small but sizable minority gave a large number.

      Another way might be to say: you've got 100 points to "spend", you must divide them somehow. That makes voter B and C more similar while leaving voter A different.

      I'm genuinely curious what people think about this.

      Whoever considers one person's life more valuable than another's will soon find himself unworthy of his own.

      by rilkas on Tue Apr 22, 2008 at 05:47:40 AM PDT

      [ Parent ]

      •  The idea of spending a certain number of points (5+ / 0-)
        Recommended by:
        theran, Wee Mama, rilkas, neroden, BYw

        has also been considered.

        But that is a lot harder, cognitively, than rating or ranking.  A lot of people would have difficulty with that.... especially in races with more than 3 candidates.  It would probably cause frustration and decrease turnout

        The nice things about range are
         1) It's familiar.  Everyone knows about a 1 to 10 scale.
         2) It doesn't force you to put one person above another, if you really don't care.
         3) It lets you indicate degree of preference.

        more in a later diary, probably next week

      •  The studies *seem* to show (2+ / 0-)
        Recommended by:
        plf515, Neon Vincent

        that a larger percentage of people vote honestly (as opposed to voting strategically) when using range voting than when using other methods.  One of its nicer features is that you can't hurt yourself too badly by voting honestly rather than strategically (with plurality voting you can hurt yourself really badly).

        If you vote strategically with range voting, you just vote the way you would with approval voting -- 100s for everyone you approved of, 0s for everyone you didn't.

        Incidentally, your example voters actually give a very good argument why range is better than ranked-choice systems, which conflate voters with very different opinions.

        -5.63, -8.10 | Impeach, Convict, Remove & Bar from Office, Arrest, Indict, Convict, Imprison!

        by neroden on Wed Apr 23, 2008 at 12:00:57 AM PDT

        [ Parent ]

      •  Accuracy (1+ / 0-)
        Recommended by:
        plf515

        One thing I would like to see before supporting that system is some psychological studies of how people apply such metrics.

        http://rangevoting.org/...
        http://rangevoting.org/...
        http://polls.zoho.com/...

        It may be easier to apply numbers than to rank candidates, but easier doesn't necessarily mean better, nor does it necessarily entail greater accuracy.

        Extensive Bayesian regret simulations say it does.
        http://rangevoting.org/...
        http://rangevoting.org/...

        With numbers you could have

        Voter A: extremist temperament -- votes 100 for favorite candidate and 3, 2, 1 for all the others...

        You can contrive inane, questionably realistic, scenarios all day.  Bayesian regret calculations actually simulate millions of elections, under every kind of electoral model imaginable (e.g. 100% honest voters, all the way to 100% strategic gaming-the-system voters) and look at the average utilitarian-ness of the results.

        Voters A , B and C would be conflated on the absolute ranking scheme, but their votes are radically different on the range scheme.

        Yes, that's the point, because intensity of preference matters:
        http://rangevoting.org/...
        http://rangevoting.org/...

        Ranked voting methods can do ever more insane things, e.g.
        http://rangevoting.org/...
        http://rangevoting.org/...

        The question is whether the range scheme more accurately portrays voter preferences or merely portrays voters' temperaments in the number assignment task.

        Question answered.  See utility figures.

        Also consider:
        http://rangevoting.org/

        In principle, the range scheme could permit a candidate whom many people gave a moderate number to, to beat a candidate who a small but sizable minority gave a large number.

        Which is good, as it increases social utility.

        •  Bayesian regret (1+ / 0-)
          Recommended by:
          plf515

          is only as good as the model.  You can run millions of simulations which "prove" range voting to be the best system, but if the model is overly simplified or otherwise doesn't apply to real life, then the results of these simulations are meaningless, even if they do seem to reproduce some important aspects of the system you are studying.  I am not arguing against range voting (I'm undecided), but your arguments are not as foolproof as you're trying to make them seem.

          Obama is the kind of politician who can get the sheep to pull the wool over their own eyes. PS - I voted for Obama.

          by Reframing the Debate on Fri Apr 25, 2008 at 05:49:35 PM PDT

          [ Parent ]

        •  All of the examples on the range voting site (1+ / 0-)
          Recommended by:
          plf515

          fail to take into account statistical error in polling.  They also fail to take into account the actions of multiple groups.   Here is a good example:

          http://rangevoting.org/...

          The example works because the voters who "exaggerate" know the outcome in advance.  It would be interesting to see what would happen if the game had incomplete information (e.g. margins of error on the polls on which the players acted).  It would also be interesting to see what happens when the other groups try to thwart Best's supporters.  I wonder if these factors (or others that I haven't thought of) wouldn't minimize gaming the system with IRV.  It would be interesting to see how "stretching the vote" in range voting such as the 40-30 becoming a 100-0 vote would affect the results.

          Obama is the kind of politician who can get the sheep to pull the wool over their own eyes. PS - I voted for Obama.

          by Reframing the Debate on Fri Apr 25, 2008 at 07:13:43 PM PDT

          [ Parent ]

Subscribe or Donate to support Daily Kos.

  • Recommended (135)
  • Community (62)
  • Elections (40)
  • 2016 (38)
  • Environment (36)
  • Bernie Sanders (36)
  • Hillary Clinton (31)
  • Culture (30)
  • Media (29)
  • Republicans (29)
  • Climate Change (27)
  • Education (23)
  • Spam (23)
  • Congress (23)
  • Civil Rights (22)
  • Trans-Pacific Partnership (21)
  • Barack Obama (21)
  • Labor (21)
  • Law (20)
  • Texas (20)
  • Click here for the mobile view of the site