[EM] Benham "zero info strategy = honesty" criterion

Chris Benham chrisjbenham at optusnet.com.au
Thu Jan 4 06:38:33 PST 2007



Warren Smith wrote:

>Range voting is criticized  for disobeying this criterion.
>Meanwhile mot rank-ballot systems like Condorcet, Borda, and IRV
>obey it (apparently; I only have proof for Borda).
>
>BUT do they?  It seems to me that in many Condorcet systems in which
>equal rankings are permitted, your best zero info vote in a situation
>with a bunch of nearly (but not exactly) equal "good" candidates,
>and another bunch of nearly-equal "bad" ones, is the rank the good ones
>co-equally (but dishonestly) top.
>
>Do you agree this is better strategy?  In which systems?  
>In particular I think this is true of Simmons' recent system
>(but his recent conversion of it from a very simple to state and understand
>system, into a mysterious one, is to be disparaged...)
>
>wds
>

In the "defeat-dropper" Condorcet methods like Schulze that use "winning 
votes" to compare defeat-strengths
the best zero-info strategy in the situation you describe is to rank all 
the "good" candidates co-equal top as you
suggest and also to strictly rank all the "bad" ones, if necessary at 
random.

But if those methods instead use Margins, then the criterion is easily 
met. Also by Schwartz,IRV.

In that situation voters in DMC have incentive to equal-rank at the top, 
but not to random-fill. The same is true of
the non-Condorcet method ER-Bucklin. All the methods that meet FBC  give 
the zero-info. "strong binary preference"
voters incentive to at least equal-rank at the top (so they tend to 
verge on being strategically equivalent to Approval).

The criterion is more-or-less met by Approval, and is easily met by IRV 
, FPP, Borda Elimination, etc.
It is failed by  Descending Solid Coalitions (DSC) which has a 
random-fill incentive, and by Descending Acquiescing
Coalitions (DAC) which has a truncation incentive.

>In particular I think this is true of Simmons' recent system..
>
No, because it is in effect based on the symmetrically completed 
ballots, since it uses "fractional" first-preferences.
It meets the criterion, as does the version I suggested, Schwartz//Simmons.

I am still enthusiastic about that version (but not the one that uses 
3-slot ballots and a "Markov process").

Chris Benham









>Range voting is criticized  for disobeying this criterion.
>Meanwhile mot rank-ballot systems lik Condorcet, Borda, and IRV
>obey it (apparently; I only have proof for Borda).
>
>BUT do they?  It seems to me that in many Condorcet systems in which
>equal rankings are permitted, your best zero info vote in a situation
>with a bunch of nearly (but not exactly) equal "good" candidates,
>and another bunch of nearly-equal "bad" ones, is the rank the good ones
>co-equally (but dishonestly) top.
>
>Do you agree this is better strategy?  In which systems?  
>In particular I think this is true of Simmons' recent system
>(but his recent conversion of it from a very simple to state and understand
>system, into a mysterious one, is to be disparaged...)
>
>wds
>http://rangevoting.org
>----
>election-methods mailing list - see http://electorama.com/em for list info
>
>  
>



More information about the Election-Methods mailing list