[EM] Ossipoff's "12 best" article

[Kristofer Munsterhjelm]

Pretty busy times here, but I can take time off to say this:

On 05/23/2017 05:35 PM, Jameson Quinn wrote:
> I think the following article contains much that's worth discussing on
> these lists
>
> https://democracychronicles.org/12-best-voting-systems/
>
> I'm going to have my own comments on this, but I'll send this out first
> to give people a chance to read it for themselves.

What the article says isn't all that surprising, given what Mike has 
written on EM itself.

Let's take the Approval CD analysis first. If you assume that:

- There are three candidates of importance
- There's an U/A setting
- Strategy is being coordinated by large party factions
- These large party factions strategize maximally,

then the conclusion follows (except for the bit "the best available 
information", as if the parties maximally strategize, then the polls may 
also be suspect).

But this isn't what concerns me about Approval. What concerns me about 
it is a semi-strategic/semi-honest situation where the C (Nader) voters 
"really" have a C>B>A (Nader>Gore>Bush) preference. Now the honest 
voters have to do all the cognitive burden of finding out whether to 
approve CB or only C, and if they get it wrong, they're not going to 
like Approval.

See Forest's remark on Approval "externalizing" strategy at 
http://lists.electorama.com/pipermail/election-methods-electorama.com/2016-October/000717.html 
and my post about manual DSV at 
http://lists.electorama.com/pipermail/election-methods-electorama.com/2016-October/000715.html.

Mike might respond that the same problem exists with ranked voting and 
so the argument above doesn't save e.g. MAM. But a honest voter knows 
how to vote in MAM: C>B>A. If the honest voter values his integrity more 
than he does getting his favorite, then ranked methods let him vote 
unambiguously. If there are no honest voters, then I agree that the 
problem also exists here, but that's a pretty big if.

Apart from that, I can see the use of CD-resistant ranked voting methods 
in situations where there's a lot of offensive strategy. In my 
experiments with a certain class of Condorcet methods, there seems to be 
basically two types of method as far as strategy resistance goes:

- Reversal symmetric, monotone methods where ~70% of IIC 3-candidate 
ballot sets are susceptible to strategy at best,
- other ones at about 10% at best,

and methods that pass the CD criterion (or suitable generalizations of 
it for voting methods where only strict full ranking is permitted) tend 
to be in the second category. I use words like "tend to" because I 
should really check this more carefully.

What are MAWP and MAWP2? I haven't heard those abbreviations before.

It'd also have been nice to have seen some mention of how well these 
methods resist strategy on the party end, such as cloning. But again, my 
impression is that I'm more concerned with cloning and Mike is more 
concerned with voter-induced strategy.