[EM] Best IRV Tweak

[Juho Laatu]

In some sense all "tweaks" are quite ok. If we want the method to be Condorcet compliant, and it is, any such method probably works quite well in practice.

I mean that in most typical large real life elections there is a Condorcet winner, and with sincere votes all Condorcet compliant methods pick the same winner. And if there is no Condorcet winner, the method probably picks a reasonably good candidate anyway. In addition to this there could be also some (rational and/or irrational) strategic voting. But also after considering these problems, differences between different Condorcet methods might still be minimal in practice.

The conclusion might thus be: any Condorcet compliant method is likely to work quite well in practical real life elections, and is likely to meet our (practical) requirements quite well. Also finer tuning is possible, but in light of what I wrote above, that might be more for artistic reasons than for practical reasons :-) . So, the question is: can we say that all Condorcet compliant methods are quite good in practice, and therefore having to "tweak" because of external requirements is quite ok? (assuming that Condorcet compliance is among our requirements in the first place)

BR, Juho



> On 8. Jul 2021, at 12.41, Kristofer Munsterhjelm <km_elmet at t-online.de> wrote:
> 
> On 7/7/21 7:47 PM, Richard Lung wrote:
>> Dear All,
>> There is science and there is politics. And I believe the latter is getting in the way of the former. And if the science was clearer, I also believe it might open up a better prospect for the politics.
> 
>> The fatal flaw of all these "tweaks" is that they are tweaks or ad hoc after-thoughts. It is that unscientific thinking which also makes the ad hoc additional member system or Mixed Member Proportional, the HOW NOT TO DO IT of election methods.
> 
> I agree that a method must live or die by the strength of its properties, and not much else. If a method appears to be intuitive but returns rotten winners, too bad for that method.
> 
> So why then Condorcet-IRV tweaks? Two reasons. The first is that they seem to be the only known methods that pass both Smith and DMTBR, which is useful when you want max resistance to strategy and still be Condorcet. And the second is that legislators used to IRV unfortunately impose a "must look not too different from IRV" criterion.
> 
> I'd wish that were not so, so that we could just advocate Ranked Pairs and be done with it. But if people are concerned about DH3 (or FairVote uses its possibility to say their method is better), then DMTBR it is. And if the politicians need simple methods - or need IRV-likes - then that's what we'll have to find.
> 
> But: that a method lives or dies by its properties is completely neutral to the origins of that method. So even a quick and dirty hack can have great property compliance. It's just less likely to be the case than if the method was properly planned ahead of time to pass the right properties.
> 
> -km
> ----
> Election-Methods mailing list - see https://electorama.com/em for list info