<DIV>Dear Jobst,</DIV>
<DIV> </DIV>
<DIV>I think you are right: Plain random ballot (as fall back) induces full cooperation at lower values of alpha than does a mixture of plain and approval random ballot, since the penalty is greater for failing to cooperate in the former case.</DIV>
<DIV> </DIV>
<DIV>However, given a value of alpha for which both fall back methods (plain and mixed) induce full cooperation, it seems to me that the fallback mixture will give a higher probability of the compromise candidate A being elected, since the probabilities only differ in case of fallback, where random ballot favors A1 and A2, while random approval ballot favors the compromise A.</DIV>
<DIV> </DIV>
<DIV>My Best,</DIV>
<DIV> </DIV>
<DIV>Forest<BR><BR>----- Original Message -----<BR>From: Jobst Heitzig <HEITZIG-J@WEB.DE><BR>Date: Tuesday, May 27, 2008 3:40 pm<BR>Subject: Re: [english 94%] Re: D(n)MAC/RB<BR>To: fsimmons@pcc.edu<BR>Cc: election-methods@lists.electorama.com<BR><BR>> Dear Forest,<BR>> <BR>> a quick calculation for your suggestion (please check!) gives:<BR>> <BR>> Winning probability for A under full cooperation of the A1 and <BR>> A2 voters:<BR>> (16+4*8)/81 + 8/27*1/2*2/3 = 56/81 = approx. 70% (OK)<BR>> <BR>> Gain in expected utility for the A1 voters when reducing their <BR>> cooperation by an infinitesimal epsilon:<BR>> epsilon*(<BR>> 4*3*(2/3)²*1/3*(<BR>> 1/2*1/3*(alpha1-alpha)<BR>> + 1/2*1/3*(alpha2-alpha)<BR>> + 2/3*(beta-alpha) )<BR>> + 8/27*1/2*(<BR>> (alpha1-alpha)<BR>> + 1/3*1/(2/3)*(alpha1-alpha) )<BR>> )<BR>> = epsilon/27*(14alpha1+8alpha2+16beta-38alpha)<BR>> <BR>> In this, alpha[1|2] and beta are the utilities of A[1|2] and B <BR>> for the <BR>> A1 voters. We may assume that alpha1=1 and beta=0.<BR>> <BR>> The A1 voters have no incentive to reduce their cooperation as <BR>> long as <BR>> the above gain is <0, i.e. when alpha>(7+4alpha2)/19. The latter <BR>> is <BR>> always true when alpha>58%. Similarly, the A2 voters will fully <BR>> cooperate when they rate A at least at 58% the way from B to <BR>> their <BR>> favourite A2.<BR>> <BR>> This is good, however with pure Random Approval as fallback it <BR>> is even <BR>> better, it seems: The gain then changes to<BR>> epsilon*(<BR>> 4*3*(2/3)²*1/3*(<BR>> 2/3*(beta-alpha) )<BR>> + 8/27*(<BR>> (alpha1-alpha)<BR>> + 1/3*1/(2/3)*(alpha1-alpha) )<BR>> )<BR>> = epsilon/27*(12alpha1+16beta-28alpha)<BR>> which is negative even when alpha>3/7 only!<BR>> <BR>> (Please double-check these calculations!)<BR>> <BR>> Yours, Jobst<BR>> <BR>> <BR>> fsimmons@pcc.edu schrieb:<BR>> > Jobst,<BR>> > <BR>> > After thinking about your recent example:<BR>> > <BR>> > > 33: A1>A>A2 >> B<BR>> > > 33: A2>A>A1 >> B<BR>> > > 33: B >> A1,A2,A<BR>> > >and the 66 A-voters try to cooperate to elect A by <BR>> unanimously approving<BR>> > >of her, then they still get A only with a low probability of 16/81<BR>> > >(approx. 20%) while A1 and A2 keep a probability of 64/243 <BR>> (approx. 25%)<BR>> > >each. A<BR>> > <BR>> > I have two ideas for incremental improvement:<BR>> > <BR>> > 1. For the fall back method, flip a coin to decide between <BR>> Random <BR>> > Ballot and Random Approval Ballot.<BR>> > <BR>> > Note that if Random Approval Ballot were used exclusively, <BR>> then there <BR>> > could be insufficient incentive for the first two factions to <BR>> give <BR>> > unanimous support to A.<BR>> > <BR>> > 2. Reduce the approval requirement from 4 of 4 to 3 out of 4 <BR>> matches. <BR>> > The the fall back method would be used less frequently, since <BR>> the 3 of 4 <BR>> > requirement is more feasible for a candidate approved on two <BR>> thirds of <BR>> > the ballots.<BR>> > <BR>> > Of course, this doesn't solve the general problem, and I'm <BR>> afraid that <BR>> > any attempt to automate these kinds of adjustments might be <BR>> vulnerable <BR>> > to manipulation by insincere ballots.<BR></DIV>