[EM] Why Clone Independence?

[Forest Simmons]

Somebody needs to talk about this ... I'm not sure I'm the right person ...
but here goes ...

First ... Plurality fails the "clone winner" criterion ... which means if
you replace the winner with two or more similar candidates, probably the
new winner will be somebody outside of the clone set ... spoiling the old
winner's chances by splitting the vote among loser clones.

This problem leads to "compromising" which means voting for "lesser evils"
instead of your favorite ... in an attempt to salvage something from the
mess.

Plurality's failure of the clone winner Criterion was a major motivation
for the voting reform movement ... including runoff voting as a way of
delaying the the compromise incentive until the time your favorite gains
enough traction to be a threat, but not enough of a threat to be worth
interfering with your compromises chances.  Suddenly you realize that if
you vote your favorite ... your compromise might get wiped out before your
favorite ... so your backup is not there to transfer your vote to.

In short, the runoff solution to compromising is an illusion ... a false
promise of"your vote will transfer to your second choice if your first
choice is eliminated."

Enter Borda ... a point system ... which takes care of the "clone winner"
problem ... at a cost of introducing a "clone loser" problem. A majority
winner can lose to a minority candidate garnering lots of points via
"teaming."  In effect, candidates with highly correlated profiles prop up
the points of one of them enough to surpass the majority winner in Borda
points. For example ...

60 A>B>C>D>E
40 B>E>D> C>A

Candidate A has a 60 percent majority of first place votes ... but loses to
B who is propped up to always getting 2nd place scores when not first.
With enough team mates the 2nd place scores are very close to first place
scores so 40 percent first place is enough for B to win.

You can see that A is buried under B's team ... so this clone Loser problem
leads to burial of the majority winner.

These clone distortions of burial and compromising can happen
unintentionally without any collusion of voting patterns among clones. If a
sizable set of candidates are significantly correlated in their
backgrounds, then they tend to form de facto clone sets ... their ranks on
the ballots will be highly correlated without conscious collusion among the
candidates or voters.

In the case of teaming this puts minorities at  a disproportionate
disadvantage ... it works against democratic diversity.

Imagine if A (in our example) is from a minority culture while B is from
the dominant culture.  Even though A has majority support the cultural
correlation among the other candidates props up B enough to over-ride the
majority will.

Clone independence is not just a theoretical nicety. It is crucial. The C
in CSSD stands for clone free ... serious voting method engineers go to
great lengths to avoid clone dependence.

It's a slap in the face to real EM scientists when some dilettante
nonchalantly tries to fob off onto a naive public a clone dependent method
... and almost gets away with it because of the subtlety of the concept.

I hope Kevin, Kristofer and others can fill in and clarify what I have left
out or otherwise unintentionally confused.

Thanks!

Forest
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20230124/3858becb/attachment.htm>