Reply to 52-alternative fwd

[Steve Eppley]

Mike O wrote:
>This is my reply to a copy of Steve's initial letter about the
>52-candidate election. Is it a postal balloting, or an
>e-mail balloting, or a meeting balloting by show-of-hands or
>paper ballot? 

Not a show of hands, since the members are scattered all over the 
country.  I'm not sure how many have email access and how many have 
web access.  Perhaps some ballots will be mailed in and others 
emailed in.

>I don't think it's necessary to eliminatate anything before the
>rankings are collected. If necessary, you could just specify that
>rankings be limited to, say, 2 alternatives, or limited to 3
>alternatives.

Would it make any sense to allow each voter to rank as many as they
want, but specify that only the top N of each ballot will be tallied
unless it's a close decision?  If so, how would "closeness" be 
determined, and what procedure would be good when it's close?

>For postal, e-mail, or paper balloting, I suggest Condorcet, with
>ranking-length limited, if necessary. But I have a 1-page BASIC
>program to do a plain Condorcet count, and  Rob L. has a
>perl program for Condorcet. My program is very short, & could
>quickly be typed into a computer, and Rob's program can, as
>I understand it, be used right where it is, at his web site.

I just checked Rob's website.  It can be used.  I'm providing more 
feedback in a separate message.

Mike posted his BASIC program a few months ago.  I haven't checked it 
out yet.  The most important issue is probably the format for the 
ballots; I hope it will read them from a file rather than require 
someone to enter them interactively.  (Ditto for Rob's perl program.)
Besides the ease at correcting typos in a file, it's also possible
for multiple people to assemble pieces of the file in parallel, then
concatenate the pieces for tallying.

Re: using a website for voting: Maybe a java-enabled user interface
could grab a file of candidates and ballots from the user's computer
and tally the ballots right on the user's computer.  This would make
the tallies independent of someone else's webserver, so even a
notebook computer not connected to the net could be used to tally.
All it would need is a java-enabled browser like Netscape or
Microsoft IE, which are free.  Other advantages of this: the same
tally software and interface could be used for stand-alone and wired
computers, and if java support really becomes ubiquitous so all
computers support it, then the tally software would also be
independent of hardware-platform.  Furthermore, java will allow 
for a graphical interface which may make voting of rankings more 
intuitive.

>If the balloting is to be by show-of-hands, then I'd suggest
>Condorcet///Approval if people aren't stubborn about circular-
>tie solution. Or BeatsAll//Approval if it's not possible to
>get agreement about circular tie solution. 

I'm not sure what these methods are.  How would they be faster to
to tally than Condorcet?  Are you again suggesting limiting the 
number of rankings per voter?  If so, why is 
Limited_Length_Condorcet///Approval better than 
N_Most_Approved///Condorcet?

>By the way, if you do pairwise in show-of-hands balloting, I
>emphasize that the 2-balloting methods Condorcet///Approval
>or BeatsAll//Approval would be better than using ordinary
>Condorcet in a show-of-hands vote.

Because of voter tactics when they can see the results of
some pairings before voting on others, I assume.

Thanks for all the suggestions.

---Steve     (Steve Eppley    seppley at alumni.caltech.edu)