[EM] MAM-d vs. The River

[Ernest Prabhakar]

Hi all,

I'm still trying to wrap my head around Jobst's 'River' method.    I 
bugged Steve Eppley into offering an opinion, which he kindly agreed to 
let me repost.

Jobst, is this a fair characterization?   Do you claim The River's sole 
virtue over other Tideman derivatives as 'easy of paper creation', or 
do you think it has other transparency advantages?

-- Ernie P.




-- Ernie P.

On Apr 16, 2004, at 4:07 PM, Steve Eppley wrote:

>>> What criteria does it satisfy?  Which does it fail?
>>
>> That's what I was trying to figure out. I was hoping you
>> (or someone) could look at it and render an opinion.
>> Basically, it seems to drop the "least relevant" defeat
>> in order to break cycles,
>
> I'm not sure that's a good description.  Consider this example:
>
>    z>x (56%)
>    y>x (55%)
>    x>a (54%)
>    a>z (53%)
>    a>y (52%)
>    y>z (51%)
>
> MAM & MAM-d would affirm z>x, y>x, x>a & y>z, electing y.
>
> Assuming I understand River correctly, it would affirm z>x,
> x>a & a>y, electing z.  To break the y>x>a>y cycle it drops
> the largest majority of this cycle, y>x.  Why call y>x "least
> relevant?"  Why is it better to affirm a>y than y>z & y>x?
> Jobst claimed River would not "unnecessarily" drop a majority,
> A>B, that could be used to directly decide that A finishes
> over B.  But River unnecessarily drops y>z, which could be
> used to directly decide that y finishes over z.
>
>> but otherwise it's like Tideman, and he claims clone
>> independence at least.
>
> Clone independence might depend on how he handles same-size
> majorities.  Unfortunately, I don't have time to check its
> criteria compliances, but I know it must fail Immunity from
> Majority Complaints since voting methods must be extremely
> similar to MAM to satisfy it.  I suspect it fails Peyton Young's
> Local Independence from Irrelevant Alternatives, have doubts
> about Monotonicity, and would like some reassurance (in
> the form of a proof) that it satisfies Minimal Defense
> and Truncation Resistance.
>
>>> The simplicity Jobst says is essential when a group is
>>> drawing the diagram interactively (by hand, I assume)
>>> doesn't seem important when tallying votes by computer,
>>> and we're living in the computer age now.
>>
>> Yes and no.  The reason I'm interested is that I think
>> Condorcet-style methods need greater transparency to be
>> acceptable to the public, so I'm always looking for cleaner
>> ways to present both the final and intermediate results.
> -snip-
>
> In presentations, we're pretty free to choose our examples.
> Is it necessary to present an example so complicated (that
> is, one that has many alternatives) that this would make
> a significant difference?
>
> --Steve