[EM] (Fwd) Re: [SWC] Outline-driven group writing

[Steve Eppley]

Here's the second of two forwards.

I'm not satisfied with the outline structure I proposed.  Now I'm
thinking a hypertext format will be needed, so the document can be
viewed organized either by electoral methods or by advantages &
disadvantages.

--Steve

------- Forwarded Message Follows -------
Subject:       Re: [SWC] Outline-driven group writing
From:          Mike Ossipoff
To:            Steve Eppley
Date:          Sat, 10 Feb 96

I have no problem with that outline structure. But, it seems to me that
rebuttals to "pros", and replies to the rebuttals, etc., should be
included in the "pros" section, so that replies & corrections could
be right after what they're replying to, rather than in a different
section.

So yes, propose that outline structure to the committee, and
we'll find out how they feel. Unless people consider it complicated
& daunting, it's fine with me.

But I suggest that methods not yet proposed could have their 
sections written at such time as someone proposes them. If someone
asks why we didn't include Copeland's method, Borda's method (Most
Valuable Player Point system), rating systems like 0-10, Bucklin,
Nanson's method, Young's method, the use of selections sets with
Condorcet's method, Smith-Borda, etc., then we can tell them that
no one proposed it, but that he/she is welcome to propose it now,
and, optionally, to write contribute to the "pro" section, and we
will write our pros & cons on it.

Yes, I like the outline form you suggested in the last message, but
I'd rather not go by categories. They're very misleading. Rank-balloting
methods are so different from each other as to not belong together
in a category, for instance.

Copeland's method is very popular with academics, but not with any
electoral reformers, and I don't like it either. A member of the
ER list proposed Smith-Borda. Some academics would want "selection
sets" to be used, but Condorcet's method us nearly always rejected
by people when that complication is added.

There are combined methods, like the Mixed Method, where rankings
are counted by all the proposed methods, and a 2nd balloting held
between the winners. And there are proposals that, whatever method
is being used, a voter can indicate on his ballot that he wants
the winner by a method _he_ specifies moved to the top of his ranking,
and the count repeated--this of course would only require 1 repetition
of the count, and would have the effect of letting the voter who
likes MPV cast a solid vote for the MPV winner, and the voter who
likes Condorcet cast a solid vote for that method's winner, regardless
of what the official count method is. I'd publicly insist on that
option if MPV were adopted for public elections.

There are other possibilities: Bucklin (which I'll define briefly
in a minute) could be modified so that the count would be repeated,
and, in the 2nd count, rankings could be shortened to the length
that elects someone as high as possible on that ranking.

In general, there could be "Revised Vote" versions of methods, where
either optionally or automatically (if most prefer), rankings could
be modified after the 1st count, so as to optimize the result in the
2nd count according to that ranking.

I'm just saying that there are a lot of possibilities, and some of
them are popular with academics or have been proposed on ER. If we
undertake to include all those methods & categories in an initial
outline, it will confuse people. Better to let people add methods
that they propose. There's nothing wrong with that simplification;
anyone on ER could propose any method to us at any time, for us to
include in the outline.

Bucklin:

Everyone's vote goes to the top candidate in their ranking.

If no one has a majority, then each ranking, additionally, gives
a vote to its 2nd place candidate.

If no one has votes exceeding half the number of voters, then each
ranking again gives a vote to its next candidate.

This continues till either a candidate has votes exceeding half the
number of voters, or all rankings have given to all their choices,
whichever happens 1st. At that time the candidate with most votes
wins.

***

Bucklin was used in several U.S. cities, earlier in this century,
including SF. I consider it the best easily-hand-counted method,
but not really a choice for public elections, now that we have
computers.

***

Copeland is a Pairwise-Count method, like Condorcet's method.

Copeland:

As in Condorcet's method, A beats B if more voters rank A over B
than vice-versa. 

Each candidate's Copeland score is calculated by subtracting the number
of candidates who beat him from the number that he beats. The candidate
with the highest Copeland score wins.

Very tie-prone, even in public elections. e.g.  With just 3
candidates, if no one beats everyone, a tie is inevitable.

Doesn't get rid of lesser-of-2-evils problem. Again, the category
of Pairwise-Count methods is misleading, since it includes Copeland
& Smith-Borda, whose merit is drastically less than that of
Condorcet's method.

***

Smith-Borda:

If no 1 candidate beats each one of the others, then, choose a
winner from the Smith set--the smallest set of candidates such
that every candidate in the set beats every candidate outside
the set. Choose from that set using Borda's method--the point 
system that gives each candidate in your ranking a point score
according to where he is in your ranking. Each rank position
is worth 1 point less than the rank position above it.

Carl Wilson proposes Smith-Borda. It's no good. Neither is Copeland's
method, though neither is as bad as MPV.

***

Some academic authors would insist that the winner always be someone
in the Smith set. Ideally a good thing, but most people consider
too complicated any method-definition that mentions the Smith set.

***

Nanson's method repeatedly eliminates the candidate with lowest
Borda score. It was used in a Wisconsin city at one time, and
still is proposed once in a while. Picks Condorcet winner if
each voter sincerely ranks all the candidates, but doesn't have
Condorcet's method's valuable lesser-of-2-evils & majority-rule
properties.

***

You get the idea: There are _lots_ of methods that have been
proposed, and which still have proponents. We don't want to
undertake to include all of them in the outline.

***

I do believe that a general dicussion of methods would be a good
thing, after the committee's task is completed. But start with
just the proposed methods, so we can get a result. And then
cover all the methods, and if that broader study uncovers a better
method then we can revise our recommendation & FAQ.

***

The "revised vote" methods is something that has only recently
occurred to me, at least as a general category. It's unquestionably
useful with some methods, but I don't know how many would be helped
by it. For instance I haven't thoroughly worked out its use in Bucklin,
but it is clearly useful in Condorcet's method, with a devious
electorate, as are several other refinements. 

My Condorcet proposal is the most basic no-frills one, because
that's the only one that gets any acceptance. It's still better
than any other proposed method that I know of, though there's always
the possibility that other good methods can turn up, such as the
possibility of a good revised-vote Bucklin method.

Obviously any proposals would be limited to precisely-defined
& completely-worked-out methods. But even with that limitation,
the number of methods is huge.

***

This, I realize, is a completely un-structured letter. That's just
occurred to me. It hadn't occurred to me till now to do structure
it.

***

In summary, those are my suggestions. 1) Categories are controversial--
and don't do a good job of separating methods of different merit.
2) There are too many methods to include them all, and I suggest just
including proposed methods, with the understanding that any SWC
member, or any ER member can propose one at any time, to be
added to the outline, after which anyone can contribute to its
pros or cons.

3) In general I agree with your recent outline propsal, & I'm
for it if others don't object.

[About revised-vote Bucklin, it may turn out to be one of those
false-leads that sometimes come up in the search for good single-
winner methods. It isn't worth discussing now, but it's the kind of
thing that someone could add when/if it's ready]

***

An interesting possible topic, after the initial task, & the
general study (if we organize things that way) would be a 
discussion & recommendation on _small elections_, as opposed
to public elections. Because of their extreme tie-proneness,
small elections (like our vote on ER list solutions) have their
own complications, mainly the need for a list of tie-breakers,
plus a random solution in case the tie-breakers fail. Not relevant
to public elections, which is our 1st concern, but nevertheless
needed _frequently_ by committees, such as SWC & ER-ADMIN. Also of
interest to committees is voting when there's no official method,
or when some don't like the official method, situations for which
I've discussed remedies in this letter.

***

Anyway, in general, I agree to the outline proposal in your latest
letter, with the suggestions that I've made for limiting it to
proposed methods.

***

I suggest the following series of possible SWC projects:

1. Discussion, recommendation, report & FAQ on the proposed methods.
2. Same, on general topic of single-winner methods for public elections.
3. Same on small elections, such as committee votes.

***

Excuse the length & non-structuredness of this letter.

***


Mike