[EM] election-methods Digest, Vol 17, Issue 46

[Rob Brown]

Warren Smith <wds <at> math.temple.edu> writes:
> Your little essay about "how political parties form" aka "movie night"
> started out nice but got lame at the end.

You mean the "happily ever after" part?  (I'll agree that was lame and was kind
of a joke.  I would like to find a better ending)

Or is there some other reason you consider it lame.

> (You also exhibit some high class knowledge of how to create web pages...
> my web pages use old technology and I think simply cannot do the
> stuff you did...)

Thank you.  I will continue to press the "presentation counts" issue here.  :)

> --First of all, CRV's range voting rules ALLOW voters to 
> "throw up their arms and say `I don't know!!!'" about a candidate.
> This is called an "intentional blank score."  If a voter does that,
> the effect on that candidate's average is exactly the
> same as if that voter had never existed.

Yes, well I personally find that answer insufficient.  I find that hardly
different than saying "hey they always have the option to stay home from the
polls and not vote at all".  

My problem is when you want to vote for a set of candidates, but are unsure what
scores to assign them. Saying "just don't vote" in no way solves that problem,
and strikes me as, frankly, a cop out. 

To make another analogy, it's like saying "you think the user interface to our
iPod knockoff is hard to use to pick a song? Just pick 'random shuffle' and your
problem is solved".

No it's not.

>> 2) Ranked ballots are already in place in San Francisco and Australia.  I
>> think it is an easier sell when a system is already making inroads. 
> 
> --It is an easier sell when you do not have to throw out your voting 
> machines and replace them with completely new system.  In San Franciso 
> in all their non-obvious IRV races, the softwar crahsed and died and 
> then everybody had to wait weeks to get the election results.  Rnage
> voting works even on today's voting machines.

Well, IRV is the problem there, not ranking.  IRV operates directly on all the
ballots, which is inherently more difficult.  Condorcet, since the "real work"
is done on the pairwise matrix, makes it easy to make the input data available
to all, and anyone with a simple program to process it can see the results.  (in
fact, right now I am working on a javascript way of doing that, so a web page
can deliver a pairwise matrix, and the actual condorcet processing of that data
is done on the client side, so that people in the know can see that it is done
right)

Anyway I'm not sure I see how range voting is more compatible with today's
voting machines, at least not the part of collecting the vote from people (i.e.
the user interface).
 
>> 2) The fact that they use IRV is mostly inconsequential, since the 
>> hard part is
>> educating people, and no need to reeducate when going from IRV to a
>> Condorcet method.
> 
> --huh?

The main thing you have to educate people about is how to rank the candidates. 
IRV and Condorcet have the exact same user interface.  In San Francisco, they
called it "ranked choice voting" anyway, so a switch from IRV to Condorcet
doesn't really need a name change.

> --If you (the voter) prefer to give a neutral score, then do so. 
> Who is stopping you?
> However, that in fact WOULD be expressing an opinion, i.e. will
> move that candidate's result away from where it would have been
> without you.  If you truly wish to express "no opinion", employ
> our option.

You haven't clearly explained why I would want to (from a strategic point of
view).  Therefore I think having the option is bad. It's a psychology thing.

>> 4) For instance, I would prefer it have the
>> scores range from -10 (strongly dislike) to 10 (strongly like), with
>> the default being zero (indifferent).  
> 
> --that is probably a mistake because it makes confusion and fraud
> more likely.
.

I don't agree or see what that is true.  As someone else here pointed out,
negative numbers map well to the concept of "dislike" and postive numbers map
well to the concept of "like".

>> 5) Unless you can figure out a way to
>> explain, in one sentence that could be understood by a third grader
>> ....."for what reason would it be in my interest to vote 'no 
>> opinion', as opposed to just picking a score midway between
>> 'like' and 'dislike'?"
> 
> --See answer to (3).  [4 sentences.]

You did not explain why it would be in my interest, as in "why would I do that
rather than ranking them neutrally?"
 
> --You are completely right - range voting will no doubt engender 
> some degree of strategic voting.  But so will essentially every 
> other voting system.  You seem to favor Condorcet systems (?); 
> many of them are especially vulnerable to strategic voting.


I don't believe Condorcet system are vulnerable to the sort of strategic voting
that would force people into opposing clusters, which range voting would (and
plurality does).  The only strategic voting in a Condorcet system would be very
subtle, to the degree it would be so unpredictable in the months leading up to
an election that it would not be a strong factor, in my opinion.  I do not see
partisanship being a result of condorcet, while I see it absolutely being a
result of range voting.

Honestly, I think that a rational person would use range voting in exactly the
way they would use approval voting....they would give each candidate either the
maximum score or the minimum score.

Have you put range voting under the a game theoretical analysis?  I'm pretty
sure the Nash equilibrium would find everyone voting this way.

> So it just comes down to which systems incentivize greater voter dishonesty,
> and which systems suffer the most from that dishonesty.  In range voting,
> for example, there is never a strategic reason to dishonestly vote your 
> favorite below top.  Ever.  But that statement is false for every Condorcet
> system and every system employing strict full rank orderings as votes.

Hmmm. I gave you real world, BASIC examples where people would vote insincerely
in range voting.  Can you give me something like that for Condorcet?  I suspect
that the only case you could would be extremely contrived, or extremely subtle
the effect insincere voting would have.

-rob