[EM] Displaying intermediate results in Condorcet-based elections

Rob Brown rob at hypermatch.com
Thu Oct 30 10:29:08 PST 2003


At 04:14 AM 10/30/2003, you wrote:
>Sorry if I wasn't clear - I have no objection to using Condorcet to select
>the winner - I applaud that and really like the voting interface. I am just
>struggling with how to make the scalar values meaningful as intermediate
>results.

And I am struggling with the same thing. :) I guess the difference is I am 
just a bit more optimistic than many as to whether or not it can be done 
effectively.

I should point out that these are not just "intermediate results", they may 
well be final results.  Even after a poll is closed, rankings and scores 
may be very useful.  In many polls and elections, being second or third 
choice does count for something, since not all elections are meant to pick 
a single winner, but might instead be used to simply measure strengths of 
opinions. For instance, say you are voting for the most annoying bugs in a 
software product, which the developers can use to prioritize bug 
fixes.  They want to know not only who is the "winner", and not only the 
ranking order of the candidates....but they want to know the relative 
*strength* of each candidate (and note that having a pairwise matrix 
doesn't help them a lot, while having a set of scalar scores does).

>Since the last ballot that comes in can change any of the victory
>strengths used in Beatpath or Ranked Pairs, the scalar values aren't useful
>as a guage of how far ahead the winner is, and this could confuse the voters
>who are used to other methods.

While I understand your point, I still believe that scalar scores can give 
an indication -- albeit not a perfect one -- of how far ahead one candidate 
is from another.  For instance, if the scores show a wide gap between two 
candidates, that's an indication that it would take a large number of votes 
for the lower ranking candidate to overtake the higher ranking one.  If the 
gap is narrow, the lower candidate could overtake the higher one with fewer 
votes.

To me, and I expect to many, that is intuitively meaningful information 
....and it would be very hard to directly deduce from a pairwise matrix, or 
from a count of all ballot combinations.

>It's not a bad thing - just depends upon the purpose of displaying
>intermediate results. As long as the results are labeled with something like
>"If the election were over now, the winners would be..." or some such
>disclaimer it would be fine.

Well I tend to think that it is implied by the fact that the poll is still 
open, but ok.

Personally i think many are looking at this a bit narrowly.....not all 
elections and polls necessarily have an "end" that is so important.  When I 
vote in an opinion poll something on CNN.com, they show me the result 
immediately, and I doubt I have ever bothered to come back "after the poll 
is closed" to see the final result.  It doesn't really matter to me.  The 
polls could stay open forever for all I care, and it would still be just as 
meaningful to view the current results (well, you can debate the overall 
meaningfulness of web polls like those on CNN, but that is a different issue).

>You just don't want a lot of after-the-fact
>questions like "How could Sally have lost? She was 'ahead' by 30 points
>yesterday...."

Hopefully people would understand that in the meantime, people came along, 
and voted for someone other than Sally -- would they not?  If they are 
confused by that, I don't know what to do.....  This is true for any 
election method: the score right now may not be the final result.  However, 
if Condorcet-based systems were really so erratic and unstable that this 
effect became extremely pronounced,  I don't think I'd want to use 
Condorcet.  Luckily I just don't think this is the case.

>I think the first few suggestions were good ones for your purposes. For the
>generalization to other purposes where the electorate understands that it is
>using Condorcet, I'd prefer something like my list of all ballots with
>non-zero votes, because that is more accurate an indicator of where my vote
>fits with the others.

Understood, although I guess you understand that in an election such as the 
recent California governor one which had 130 candidates, that would be a 
whole lot of ballot combinations for you to look through.

But indeed, if this thing I'm doing is a success, I'm sure I'll eventually 
have an option for some sort of advanced view where you can see all that 
kind information if you want.  But being as my goal is to get as many 
people as possible to embrace Condorcet-based methods over plurality, I 
think it is essential to produce simple, straightforward, clearly 
digestible graphic output -- which to me means scores that can be put in a 
bar graph.

-rob




More information about the Election-Methods mailing list