[EM] Re: Total Approval Ranked Pairs

[Ted Stern]

On 15 Mar 2005 at 21:49 PST, Russ Paielli wrote:
> I'm just trying strike a balance between simplicity and
> effectiveness. I am starting to realize that equal rankings may be
> worthwhile. As for allowing ranking past the Approval cutoff point,
> I am still not sold on that, but I am open minded. Think of how much
> simpler the voter interface will be if it isn't allowed. If we're
> going to recommend allowing it, we had better be sure it is worth
> the added "cost." By "cost," I mean not only the added complexity
> itself, but also the possibility that such added complexity will
> push the method past the threshold of public acceptability. Note
> that the simple idea of ranking candidates will stress the limits of
> public acceptability all by itself.

I heartily agree that it could get complex.

Chris Benham sent a message earlier, directing our attention to this
2002 post by Adam Tarr:

http://lists.electorama.com/pipermail/election-methods-electorama.com/2002-March/007699.html

I happen to like graded ballots.  In thinking about it, I came up with
two ballot formats:

,----[ seeded-bubblesort-ballot ]
| Idea one: Graded ballots.  A/B/C get an approval point for purposes of
| the initial ranking, D/E/F get none.  Unranked candidates are below F.
| 
|                Pass             Fail
|             A    B    C      D    E    F
| 
|       X1   ( )  ( )  ( )    ( )  ( )  ( )
| 
|       X2   ( )  ( )  ( )    ( )  ( )  ( )
| 
|       X3   ( )  ( )  ( )    ( )  ( )  ( )
| 
|       X3   ( )  ( )  ( )    ( )  ( )  ( )
| 
| Most people would be able to "get" this.  After all, in school an F
| usually means you were actually there to take the test.  No grade at
| all means you didn't even write your name down.
| 
| Passing votes get a point, used for the initial ranking before the
| bubblesort.  Failing grades don't get any point, but even a failed
| candidate can bubble up if it defeats all the other candidates
| pairwise.
| 
| Idea two:  9 ranks, adjustable first-sort point levels.
| 
|             Default 100       Default 50       Default 0
| 
|             A    B    C      D    E    F      G    H    I
| 
|       X1   ( )  ( )  ( )    ( )  ( )  ( )    ( )  ( )  ( )
| 
|       X2   ( )  ( )  ( )    ( )  ( )  ( )    ( )  ( )  ( )
| 
|       X3   ( )  ( )  ( )    ( )  ( )  ( )    ( )  ( )  ( )
| 
|       X3   ( )  ( )  ( )    ( )  ( )  ( )    ( )  ( )  ( )
| 
| 
| If you want, you can enter a different point level for any of the 3
| rank sets.  E.g., you could set the ABC group to 70 points, the DEF
| group to 40 points, and the GHI group to 10 points.
| 
| You'd sum up the points for each candidate into its corresponding
| pairwise matrix diagonal element.  Those point totals would be used
| for the first round of sorting.
| 
| Ideally, I'd like to be able to adjust the point level for each
| candidate separately, but this seems like the most compact way to do
| it.
`----

>
>> Another key argument about combining Instant Round-Robin (aka
>> Condorcet) with Approval should be checks and balances.  Condorcet
>> alone, Approval alone, or IRV alone can each fail in some
>> situations.  But Condorcet augmented by Approval is actually
>> simpler and more resistant to manipulation than Condorcet alone.
>
> Absolutely. More on that later.
>
> By the way, I'm not too crazy about the name "Approval Runoff
> Condorcet" (ARC). I think "Condorcet with Approval Runoff" (CAR)
> would be preferable, but I'm not too keen on that name either. Let's
> think about this some more.
>
> --Russ

Russ, do you agree with Forest, Jobst and myself that your proposal
(if ranking below the cutoff is allowed) is equivalent to sorting
twice, first in descending order of approval (or cardinal ratings),
then by bubble sort using pairwise comparisons?

Even though Paul Kislanko would disagree with the sports comparison, I
think calling this method Tournament Voting is very descriptive of the
process, and would be convey some intuitive understanding to voters of
how it works.

Specific variations might be denoted TV(Approval), TV(CR), to specify
how you're doing the initial sort.

And yes, "Tournament" is an overloaded term.  But so is "Instant
Runoff" ;-).

Just to be clear -- in taxonomy terms, IRR is the family, TV is the
genus, Approval or Cardinal Ratings is the species.

Ted
-- 
Send real replies to
	ted stern at u dot washington dot edu

Frango ut patefaciam -- I break so that I may reveal