Minority Opinions

Not everyone can be mainstream, after all.

Voting System Iceberg

with 3 comments

The more I learn about something, the more I become aware that I’ve just scratched the surface.  This week, it’s voting systems again, courtesy of Henrik Ingo’s blog.  He also writes about MySQL, which is how I discovered his writings, but what really intrigues me is his work on a secure online voting system, and his attempts to make it ideal.  Sadly, just as Arrow shot down the idea of a perfect ranked vote aggregation system, Ingo has produced a list of incompatible requirements for voting securely.  How can I verify that my vote was processed accurately, without being able to prove how I voted?

Math may come to the rescue here, in the form of encryption algorithms that allow certain mathematical operations to be performed on a set of encrypted ballots.  The explanations get a little more technical than I’m comfortable with, but if such a scheme is practical, then I can see how being able to verifiably add ballots would suffice for plurality, approval, or range voting.  What I don’t see is how you can properly format the ballot, prove that it’s properly formatted, verify that your ballot is among the ones in the final tally, and still be unable to prove to another person that you voted in a certain way.  There may well be ways to do that, particularly with certain constraints, but I’m not going to be the one designing them.

After all, I still have to convince myself that range voting is better than a Condorcet method.  I may yet get there, and I certainly buy the argument that approval voting is a perfect first step for a political body seeking improvement, but something still feels off.  Maybe it’s the way it feels too similar to plurality or a Borda count, with their obvious flaws.


Written by eswald

29 Jan 2013 at 9:51 pm

Posted in Politics, Technology

3 Responses

Subscribe to comments with RSS.

  1. Score Voting (aka Range Voting) is just objectively better than Condorcet systems per extensive Bayesiean regret calculations:

    It’s mathematically proven that the Condorcet winner is not necessarily the favorite candidate of the group. But even if you irrationally insist on always electing Condorcet winners when they exist, Score Voting may be a better Condorcet method than real Condorcet methods.

    Also, Score Voting and Approval Voting are radically simpler and more politically viable than any Condorcet method. You will *never* see Condorcet adopted in US elections. I’ve spent years trying to convince elected officials to try Approval Voting and Score Voting, and even this has been fruitless thus far.

    Clay Shentrup

    30 Jan 2013 at 1:18 am

    • So I’ve read, but I’d like to run some simulations myself. Bayesian regret certainly feels like a good objective measuring stick, though.

      It could also be that my vague distrust comes from attempting to apply voting methods to a very different problem domain, with a couple dozen candidates and a few dozen honest voters, where each voter only has an opinion on up to about seven candidates. The resulting preference lists feel like they naturally fit ranked voting systems, whereas range voting might have issues when voters aren’t using the same rating scale. That doesn’t seem to stop Amazon or other five-star rating systems, though, so perhaps my fear is unfounded. I can also see how millions of voters for a handful of candidates would wash away the problem.


      30 Jan 2013 at 10:24 am

      • The “same scale” argument is an argument FOR Score Voting, because it discards LESS information than a ranked ballot.

        Example utilities:

        Bob: X=20, Y=18, Z=10
        Alice: X=10, Y=9, Z=0

        If using, for example, a 0-10 Score Voting ballot, these utilities BOTH normalize to X=10, Y=9, Z=0. This discards a lot of information, but it’s still evident that Bob prefers X to Y by a much smaller amount than he prefers Y to Z.

        With ranking, even THAT info would be discarded, producing X>Y>Z for both Bob and Alice.

        Incidentally, the Bayesian regret figures I cited ALREADY normalize the actual utilities to the set of allowed scores, producing the very distortion you’re talking about when you complain about voters having “different scales”. Nevertheless, Score Voting still shines.

        Clay Shentrup

        30 Jan 2013 at 2:57 pm

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s