# [EM] Gradual Information Approval

Paul Kislanko kislanko at airmail.net
Fri Sep 30 16:59:03 PDT 2005

```"
A>>>B>>>>C>>>D>E>F>>G

does not imply that A is preferred over B with three times the intensity of
the D over E preference.   It only means that the strongest preference is B
over C, then next strongest is A over B, which is about the same as the C
over D  preference, etc."

I don't want a ballot that makes me figure out how many >'s to put in, so
interpreting the number of '>'s will never come up. And any method whose
definition includes "approval cutoff" is obviously not one I'd ever consider
using.

Seriously, folks this is getting more incomprehensible with every variation.

As a voter using a ranked-ballot my "approval cutoff" occurs when I stop
ranking candidates. There's nothing to figure out about my ballot and
there's no difference between > and >>> and >>>>>.in a Condorcet method, and
I'm not going to any extra work on the voting side just to suppirt an
extremely complicated method that doesn't offer any advantages.

_____

From: election-methods-electorama.com-bounces at electorama.com
[mailto:election-methods-electorama.com-bounces at electorama.com] On Behalf Of
Simmons, Forest
Sent: Friday, September 30, 2005 6:33 PM
To: election-methods-electorama.com at electorama.com
Subject: [EM] Gradual Information Approval

A while back Kevin came up with an interesting method that he called Gradual
Information Approval.

There are both ranked ballot and ratings versions of this method.

If there are K candidates then K-1 approval rounds are simulated.

In round m, the 1+K-m most approved candidates in the previous round are
given equal weight in determining the approval cutoffs for the next round.

In the ratings version, above mean strategy is used.

In the ordinal version, above median strategy is used.

What if we used ballots that are intermediate between ratings and rankings?

I refer to what I call "Ranked Strength Voting Preferences."

For these RSVP ballots you don't have to have any absolute strengths for the
intensity of preferences.  But you are able to indicate which of two
preferences you consider stronger, if you feel that one is stronger than
another.

So the ballot

A>>>B>>>>C>>>D>E>F>>G

does not imply that A is preferred over B with three times the intensity of
the D over E preference.   It only means that the strongest preference is B
over C, then next strongest is A over B, which is about the same as the C
over D  preference, etc.

I say that RSVP ballots are somewhere between Ratings and Rankings because
Rankings can be inferred from RSVP, which in turn can be inferred from high
resolution Ratings.

Now, how would RSVP ballots be used for Gradual Information Approval?

If every preference has a different strength, then the approval cutoff would
be the highest strength preference that is straddled by the set of 1+K-m
most approved candidates  from the previous round.

If there are no distinctions in strength of preferences straddled, then
we're back to the above median strategy.

Otherwise, we use a hybrid of these two strategies, using above median for
the highest strength preferences straddled by the previous step approval
standouts, etc.

Kevin, did Gradual Information Approval turn out to be monotone?  Probably
not.

What other applications would be good for these RSVP ballots?  [I mean
besides invitations to parties.]

Forest

-------------- next part --------------
A non-text attachment was scrubbed...
Name: winmail.dat
Type: application/ms-tnef
Size: 7654 bytes
Desc: not available
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20050930/07274df7/attachment-0003.bin>
```