[EM] Helping a candidate in the case of ties

[Dave Ketchum]

This is more detailed work, depending on precise knowledge.

Also gets tricky since many changes affect more than one pair of  
candidates.

For a simple example where helping A in A vs B, without disturbing  
their relationship to other candidates, will help C (could be hoping  
to cause A to be bigger than B in their pair; could be simply to  
change the magnitude of their difference).
      Tell those who would do A=B or B>A, to vote A>B.  This will  
affect A vs B without affecting any other pair of candidates.

Note that adding one or both of these, or giving them adjacent ranks  
when they had not had this, requires more complex analysis.

Dave Ketchum

On Nov 24, 2009, at 11:51 AM, Kristofer Munsterhjelm wrote:

> It's fairly straightforward to define whether a candidate is helped  
> after a change of ballots if "helping" is limited to win/not win: if  
> the candidate wasn't in the set of winners (ranked first on the  
> social ordering), but is after the modification, the candidate was  
> helped. It is also not that difficult to define it for a social  
> ordering without ties: if the candidate moves from qth place to pth  
> place, p < q, then he was helped.
>
> But how would one define this for an ordering with ties? The problem  
> with defining it in terms of candidates higher ranked is that if
> A > B > C > D = E turns into A > B = E > C > D, C is "helped"  
> according to that metric, even though intuitively it seems like he's  
> not so. On the other hand, defining it in terms of ranks above the  
> set containing the candidate has problems when the possible number  
> of sets change. For instance, A > B > C > D turning into A = B = D >  
> C doesn't seem to have "helped" C, although now he's second, whereas  
> before the change, he was third.
>
> Is there any consistent way of defning help and harm, in the context  
> of candidates, when the social ordering may contain ties?