IMC, I2C and LIIA criteria (was Re: [EM] Markus: RP & BeatpathWinner/CSSD)

[Markus Schulze]

Dear Steve,

you wrote (11 March 2003):
> Markus Schulze wrote (11 March 2003):
> > IMC, I2C, and LIIA sound interesting. But do they really say anything
> > about how often an additional candidate changes the result of the
> > elections? Or are they just curiosities? 
>
> That's a false dichotomy.  These three criteria are not about adding an
> additional candidate.  But that does not imply they are just curiosities.
> They are "consistency" conditions whose violation may disturb some
> voters, perhaps causing them to change the voting method, or perhaps
> undermining the mandate of the winner.

As far as I have understood Young correctly, his LIIA is intended to
be a weakening of Arrow's IIA. Therefore, in my opinion, it makes sense
to ask whether in simulations compliance with LIIA really leads to fewer
violations of IIA.

You wrote (11 March 2003):
> Markus Schulze wrote (11 March 2003):
> > In so far as you have already programmed some simulations, I would like
> > to ask you to make some simulations where the number of candidates is
> > increased from N to N+1. I guess that the numbers of situations where
> > the additional candidate changes the winner without being elected won't
> > differ significantly for the two Condorcet methods.
>
> I don't know if I'll have time, but in case I do, can you provide
> a clearer specification of the problem?  Should we assume the
> additional candidate is inserted randomly into each voter's ranking?
> If so, why?

The additional candidate should be inserted randomly because random
simulations have been accepted more or less as a feasible tool.

You wrote (11 March 2003):
> Is the purpose of this simulation to test which method satisfies
> Arrow's IIA as much as possible?

Yes.

Markus Schulze