On Fri, Dec 06, 2002 at 03:13:59PM +1100, Clinton Mead wrote: > First, heres a definition of a rough proposal I've used in examples > below, which I've called, for want of a better name, Considered CSSD > (CCSSD). Its a clean up and patch of a similar earlier proposal.
I basically have a lot of questions, starting with: What do the other letters stand for (CSSD?). I ask, because the system you propose is further from pure Clone-proof Schwartz Sequential Dropping than mine is. > >D defeats A 4:3 > >A defeats B 4:1 > >B defeats D 4:1 > > > >eliminate 4:1 > > > >D defeats A 4:3 > > > >tie between B and D > >______________________________________________________________________ > > > > > CCSSD: B wins. > > Here, the B voters are penalised for being sincere. Actually, my proposal favors B voters -- in a pure CpSSD system (without supermajority requirements), A would have won. > The B voters as a whole prefer A over B, and being sincere voters, > propose A and vote sincirely for it. If the B voters were insincere, > by voting B over A, they would of recieved a more preferable result. Why do you say that "B wins" is more preferable to "tie between B and D"? If the person with the casting vote chooses "B", the outcome is the same. If the person with the casting vote chooses "D", we continue with further discussion on how to pick a better option [perhaps "A" will win in the next election, perhaps "B" will win, or perhaps some better option "C" will win]. > >A requires 2:1 majority; D is the default option > >60 ABD > >30 BAD > >10 DBA > > >B defeats D 90:10 > >A defeats D 90:20 > >A defeats B 60:40 > > > >A wins > >______________________________________________________________________ > > > > > CCSSD: B Wins > > This is more a matter of opinion on my behalf, but I think in this case, > A should have to superdefeat both B and D to be declared winner. Why? The most fundamental difference between your proposal and mine [the one illustrated in these examples] is that yours uses the supermajority ratio in comparisons between options other than the default option. I'm not saying that this is intrinsically bad (I'll let Anthony Towns do that), but you've not really justified why this approach is better. > >A requires 2:1 majority; N is the default option > >4 cAbN > >1 cNAb > >3 bcNA > >3 AbcN > > > >N defeats A 8:7 > >c defeats N 11:0 > >b defeats N 10:1 > >A defeats b 8:3 > >c defeats A 8:3 > >b defeats c 6:5 > > > >eliminate 6:5 > > > >N defeats A 8:7 > >c defeats N 11:0 > >b defeats N 10:1 > >A defeats b 8:3 > >c defeats A 8:3 > > > >c wins > >______________________________________________________________________ > > > CCSSD: B wins > > In this vote, the 3 AbcN voters could insincerly vote bAcN. So? This is a losing strategy. With CpSSD, voting as your first preference "the lesser of two evils" only makes a difference if your real first preference has a good chance of winning. > In conclusion, I've got a vibe that treating the default option > differently to other non-supermajority options regarding supermajority > defeats can reduce stability and encourage insincere voting. What do you mean here? Would you say that a "voting system" where the default option always wins is more stable, or less stable, than the system I've proposed? Also, the only "insincerity" I've seen you propose is the losing strategy of ranking your first preference lower than first because your first preference has such a good chance of winning that it will influence the vote in a fashion you "don't approve of". I'll grant that you can present specific instances where this seems to be worthwhile, but I think that to do anything meaningful with this would require complete omniscience about how people are going to actually vote. > Though nothing hard and fast though. Point out places where CCSSD does > worse than the current draft, in the above examples or others, because > I've probably got a biased view of my own proposal. Is the above good enough? Let's start with a justification of why this approach to supermajority handling is better. Thanks, -- Raul -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
