Kristofer, I don't know about Condorcet auditing because I haven't tried to figure it out yet, but its mathematics seems to be much simpler than IRV/STV since only n(n-1) counts are necessary to report for each audit unit (precinct or whatever.) and I'm sure the fact that Condorcet counts fit so nicely into an nxn matrix would probably help the numeric algorithms.
I would need to figure out how what the upper margin error bounds for all candidate pairs for one Condorcet audit unit (precinct or other publicly reported) matrix are, given the reported vote counts and number of ballots cast in each. Once that is done, I think figuring out several methods would come simply. It seems very do-able to me, but it would take me who knows how many days or weeks or even months of studying the problem to figure out exactly. As soon as Condorcet methods are adopted for a public federal, or perhaps even state-level election, I would definitely not oppose implementation of the Condorcet method with the minimum margin method of resolving cycles and would be happy to try to develop the post-election auditing mathematics for it. Kathy On Sat, Jan 16, 2010 at 4:55 PM, Kristofer Munsterhjelm <[email protected]> wrote: > Kathy Dopp wrote: >> >> On Sat, Jan 16, 2010 at 2:54 PM, Abd ul-Rahman Lomax >> <[email protected]> wrote: >>> >>> results. Generally, voting security people like to use audits that select >>> a >>> sample of votes and look for errors, then they use statistical analysis >>> to >>> estimate the overall error in result probability. That's not nearly as >>> easy >>> with IRV, because IRV is a chaotic method, sensitive to a single vote >>> error >>> that ripples into shifting many votes. With many places where that single >>> vote error can occur. >>> >>> At least that's my understanding, I'd defer to Kathy on this! >> >> Yes, IRV is virtually impossible (practically) to audit in a way that >> the general public could understand. There are several ways to >> manually audit IRV/STV as far as publicly reporting the tallies and >> randomly selecting them: >> >> 1. report every rank choice ballot for every voter and make a humanly >> readable mark on every ballot (preferrably after the voter casts or as >> the voter casts the ballot to avoid vote-buying) that is also listed >> alongside the ballot choices, and then randomly select ballots, or >> >> 2. publicly report all the tallies for each possible unique rank >> choice vote for each precinct (a huge number larger than the number of >> voters who vote in most precincts if there are many candidates >> running), and then randomly sample from those, or >> >> 3. manually count 100% of the ballots > > Offtopic, perhaps, but would these problems hold for Condorcet as well? Both > IRV and Condorcet methods are ranked-ballot ones, though I guess auditing > Condorcet would be easier since it's precinct-summable. It doesn't appear to > be as easy as Plurality, though, because you can't tie "A beat B N times" to > what kind of votes the N voters submitted other than that they ranked A > ahead of B. > -- Kathy Dopp Town of Colonie, NY 12304 phone 518-952-4030 cell 518-505-0220 http://utahcountvotes.org http://electionmathematics.org http://kathydopp.com/serendipity/ Realities Mar Instant Runoff Voting http://electionmathematics.org/ucvAnalysis/US/RCV-IRV/InstantRunoffVotingFlaws.pdf Voters Have Reason to Worry http://utahcountvotes.org/UT/UtahCountVotes-ThadHall-Response.pdf Checking election outcome accuracy --- Post-election audit sampling http://electionmathematics.org/em-audits/US/PEAuditSamplingMethods.pdf ---- Election-Methods mailing list - see http://electorama.com/em for list info
