[EM] Intro to list (etc)

[Rob Brown]

At 01:00 PM 10/26/2003, Paul Kislanko wrote:
>As I said above, not seeing the others' votes is only important in
>elections. In the ezboard-type environment (I am a user of ezboards, and for
>sure their polls need some help - as they are today they are pointless and
>worse than useles....) But these are opinion polls, and as it happens the
>area you're working in is closely related to some work I was doing for my
>company before downsizing left the intellectual property rights with them
>and me unemployed :-(

So you're an ezboard user, huh?  Well, cool, looks like there actually ARE 
some tech savvy ezboard users out there. :)

BTW, this will likely be used for some elections, although maybe not very 
"important" ones.  For instance, people in a message board would be able to 
elect moderators for their board.  The benefits of a Condorcet style 
election is actually really important in such a context, since it can tend 
to elect "middle ground" candidates (message boards can certainly get very 
politicized and polarized!).

However, I don't think it should be a problem that people can see results 
during the voting, even in an election vs. just an opinion poll.  Since 
Condorcet (mostly?) eliminates strategy, I don't see a problem.

> >If given a choice between a standard plurality poll that shows current
> >results as soon as you vote, vs. one that does Condorcet style ranking, but
> >you have to wait until the end to see results....this type of user will
> >choose the plurality one any day.  (at least that is my prediction)
>
>This type of user (actually ALL users of poll results) want to see a
>numerical score as the results. The score doesn't have to be "plurality" as
>in only the number of first place votes, it just has to be a number. You can
>use any function you want and they'll be happy (i.e. see the modified Borda
>scoring for the weekly Associated Press and ESPN "polls" in the major
>college sports).

Yeah, but.....what if the point results don't match with the actual 
ranking?  I am pretty committed to using Condorcet to actually rank them, 
because I don't want strategy to play a part.  If I use Condorcet for the 
ranking and Borda for the graph.....you could end up with something that 
would confuse people ("why did X win when Y has more points?")

> >My little agenda here is to see if I can make a mainstream audience
> >comfortable with Condorcet style voting....hopefully people who use it for
> >little meaningless web polls and realize its benefits, people will warm up
> >to it for "real" elections.
>
>Just don't call it "Condorcet style voting" - there isn't a voter alive who
>isn't capable of listing their preferences - Chocolate, Vanilla, Strawberry
>or Rocky Road, Cookies & Cream; Oklahoma, Miami, Southern Cal, etc. Leave
>the jargon out and just let them put a list in the order they prefer.

Yeah, I won't, although I will probably link to an in-depth article 
explaining it to those who really care (such as those who will want to 
question its validity...there are sure to be ones to do that!! ).  But for 
most people, they won't know or care what it is called or care about the 
details of how it works.

>...
>Now, that's clumsier than yours, but I am limited as to what I can do on
>RIvals.com sites.

Yeah, well mine of course relies on newer browser features, and its *just* 
getting to the point where its safe to assume that the great majority of 
the audience has a browser that will support the dynamic html stuff.  (and 
even then, I'll probably have to have it degrade gracefully for those using 
ancient or text-only browsers)

>Again, for what you're doing you need only pick one vote-counting method and
>present just the results for it. My suggestion that you support multiple was
>based upon the use of the tool to analyze voting methods, not the highly
>more practical one you have in mind.  Bar graphs for displaying results to
>the ezboard members (voters or not) would be just fine, except...
>
>What would you use to construct a bar graph from a pairwise matrix? Quite
>often the condorcet winner does not have the highest A?B comparison in the
>pairwise matrix. I think you can use ranked ballots to develop a "score",
>but you can't use Condorcet to the order the results because Condorcet
>doesn't order ranks 2-n...

Yep, quite a dilemma.....  :(

>Yes. For this purpose I think Borda would work just fine. Your constituency
>is used to it, and the fate of the free world is not in the balance, so the
>fact that it doesn't meet all the EM criteria we'd like is not relevant.

Nooooooooooooooo!!!!!!   Please, not Borda!!!!   :)

I'm really REALLY hoping I can find a good way to do Condorcet.  In my 
little fantasy world, the fate of the free world IS in the balance.....see, 
if ezboard people start doing condorcet elections and like it, and then 
other sites see that it works well and copy it, and then people start doing 
it for more serious elections (say, voting for officers of an 
organization), and then governments start trying it for 
elections........well you get the idea, the free world is saved!!  ;) Yes, 
I'm being a bit facetious and grandiouse, obviously, but I really believe 
that elections should be free from strategy, so if I was to have to go with 
Borda would be pretty much admitting defeat -- and wasting a good 
opportunity to promote strategy-free voting to the masses.

I doubt ezboard cares whether it uses Borda or Condorcet, but, you know, I 
do. :)

>It's also easy to calculate and translate into the one-dimensional space
>that people can handle easily.

Well, no question Borda is easier to calculate!

But....as for translating into 1-d....doesn't Condorcet *already* have to 
translate its two dimensional table into a 1-d space?  I mean, you've got 
to pick a winner (and by extension, you can also use the same logic to 
produce a final ranking order for all candidates).  So, since the problem 
of ordering them into a one-dimensional arrangement is solved, it seems 
that taking that one step further and assigning a numerical score to each 
candidate (while maintaining the Condorcet logic) is solvable too.  Maybe??

-rob