[EM] Re: The new poll

MIKE OSSIPOFF nkklrp at hotmail.com
Mon May 24 20:48:02 PDT 2004


Brian Olson:

You wrote:

Ratings because they are most expressive, and rankings, approval or
even a single vote could be derived from them.

I reply:

No, Approval ballots can't be inferred from ratings, unless we assume, for 
instance the ratings are sincere, and that everyone is using the above-mean 
strategy for 0-info elections. Of that requires that the election be 0-info. 
The poll is not 0-info. For instance, my own Approval ballot for that poll 
is based on the alternatives' popularity, their likelihood of being top-2 
contenders.

So, for one thing, it isn't valid to assume that people would use 0-info 
strategy for a poll that isn't 0-info.

And you can't reliably assume that everyone's ratings are sincere. Without 
that assumption, the above-mean strategy assumption would be of no use, even 
if it were a 0-info election.

You continued:

I discount the importance of cheating a system with a strategic vote
because there are systems that don't reward such behavior and an honest
vote is most likely to get the outcome a voter desires.

I reply:

Ok, that's something that is news to all of us here.  Would you mind telling 
us what that no-strategy method is? Random Ballot? But you didn't mention RB 
as one of the counting methods that you will use.

Because you haven't heard about this yet, let me be the one to tell you that 
Gibbard & Satterthwaite demonstrated that no nonprobabilistic method is 
always free of incentive to vote other than sincerely. I don't know the 
wording of what they proved, but that's basically it.

Or at least,
it's my hope that we'll find one.

I reply:

No one will find one. It's been shown that none exist, except for 
probabilistic methods like Random Ballot.

You continued:

I could publish a list of anonymized votes. We could all scrutinize
them to see if one of them looks like someone was trying to be clever.

I reply:

Of course, why not post them.

I'd said:

>And I suggest that voters in Brian's poll post Approval ballots to EM, 
>among the alternatives in Brian's poll.

You replied:

I expect I beat you to it:
'c' for checkbox, Approval ballot

I reply:

Ok, I didn't know that there was a way to cast Approval votes at your 
website, for the poll.

So, I'll run the same election methods on the back end no matter what
the ballot style. IRNR, Condorcet (+beatpath), IRV, Borda, and Raw
Cardinal Rating summation.

I reply:

But Approval requires Approval balloting. If you assume how people would 
vote in Approval based on their ratings, you're just guessing.

You continued:

'emc' could be interesting. I've never run
Condorcet of an Approval ballot. It is of course possible, each
Approved candidate is preferred to each non-approved candidate,
approved aren't preferred over other approved and non-approved aren't
preferred over other non-approved. Could be interesting.

I reply:

In an Approval balloting, the candidate who gets the most votes is the voted 
CW. Compared to each one of the other candidates, s/he is voted over hir by 
more voters than vice-versa. Where, in Approval, voting X over Y means 
voting for X and not for Y.

By the way, you ask for ratings from -10 to 10. For the sincere voter, that 
balloting asks the voter which alternatives they consider to have positive 
or negative value. That's an artificial notiion to impose on these choices. 
It would be better to just ask for ratings from 0 to 10 or from 0 to 100, 
etc.

Sure, a person could transform the ratings as people have discussed here, so 
that 0 becomes -10 and 100 becomes 10. But that wouldn't be answering the 
ballot's question, about the alternatives' negativeness or positiveness.

I was saying before that we here consider the lowest allowed rating to 
represent the worst candidate and the highest allowed rating to represent 
the best candidate. But your +/1 ballot asks another question, about how 
negative or positive an alternative is. Compared to what? The midrange? Did 
you say that? Asking for positive ratings with 0 as the minimum would have 
been better.

Mike Ossipoff

_________________________________________________________________
MSN Toolbar provides one-click access to Hotmail from any Web page – FREE 
download! http://toolbar.msn.click-url.com/go/onm00200413ave/direct/01/




More information about the Election-Methods mailing list