[EM] A problem with IRV3/AV3

[Jameson Quinn]

Imagine a scenario of an ABCD one-dimensional continuum:

41: A>B>C
19: B>A>C
20: B>C>D
20: C>B>D

If the A voters vote A>>D then A will win. By raising the turkey D over the
true CW B, they have stolen the win. Even if their strategy fails to keep B
out of the top 3, they lose nothing; B will still win.

To be honest, it was harder to tune this scenario than I thought it would
be. Thus, having taken the time to write this down, I am no longer opposed
to IRV3/AV3. (For IRV2/AV2, it's easier to get this problem. It's also
easier to get the problem if there are clones involved, but real-world
clones beyond 3 candidates are unlikely.)

Since I'm now not opposed to IRV3/AV3, I consider it one of the 3 reforms
(along with SODA and IRV) that would be most acceptable to incumbents,
because it avoids the weak Condorcet winner problem. Still, it is basically
just as bad as IRV for nonmonotonicity and spoilers; all the spoiler
scenarios I consider realistic are essentially 3-candidate anyway. As such,
I see no reason to believe that it would not lead to lesser-evil voting and
2-party domination, as IRV does. Since I see 2-party domination (as opposed
to just having 2 strongest parties, a logical necessity) as a source of the
most-serious problems with Plurality, I still feel that SODA is a much
better option than IRV3/AV3.

Jameson
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20120108/48110402/attachment.htm>