[EM] STAR

[Kristofer Munsterhjelm]

On 8/19/23 05:10, C.Benham wrote:
> Toby,
> 
>> I wouldn't count this as a monotonicity failure because it involves 
>> decreasing Y's score as well as increasing X's.
> 
> 
> This is like a sophist's technical loophole.  Why do we particularly 
> care about "monotonicity failure"? To avoid some hypothetical mild 
> embarrassment?
> For the sake of marketing bragging rights?
> 
> Or because it is related to Push-over strategy
> incentive/vulnerability? STAR is  much worse in that respect than IRV
> because there the strategists are entirely relying on other voters to
> both get their favourite into the final two and to there win the
> pairwise contest, so if too many of X's supporters try the strategy
> it could backfire.
I'd say there are two problems when a method shows nonmonotonicity. The 
first is pushover incentive, and the second is that the method somehow 
is contradicting its own judgement, which makes that judgement suspect.

Since no method can be consistent in every way possible, it's not 
possible for its judgement to be free of flaws in the second sense. But 
as we have monotone methods, it seems this particular flaw can be 
avoided. (Well, unless we want both LNHs and mutual majority.)

As an example, consider my Methd X of earlier. My simulations seem to 
indicate that it has no pushover incentive at all, i.e. if A won, it's 
impossible for B>A voters to make B win by upranking A or downranking B. 
However, it still has a problem where B>A voters upranking A can change 
the winner from A to C. If monotonicity were all about pushover 
resistance, I'd be celebrating now and say the thing I've been trying to 
solve for years is done. But it's still not entirely monotone.

Or Rob Richie has responded to objections that IRV is non-monotone with 
"but you need too much knowledge of the election to pull off pushover, 
so it's not a problem". But it's still not monotone.

-km