[EM] eBooks - naming - MAM-d

[Eric Gorr]

(Posted with Ernest's permission)

At 9:30 AM -0800 2/4/04, Ernest Prabhakar wrote:
>On Feb 4, 2004, at 8:59 AM, Eric Gorr wrote:
>>Considering that this method is (as near as I can tell) identical 
>>to MAM in every way, but one. The way it differs from MAM is that 
>>it does not contain any random selection procedure when faced with 
>>equal defeats, which are apart of a cycle, choosing instead to 
>>reject those defeats - which is why the word 'deterministic' is 
>>used.
>>
>>So, for a nice short name, it could be called MAM-d ... where the 
>>'d' stands for  'deterministic'.
>
>Interesting.   I'm curious how such rejection affects things like 
>clone-independence.

It is clone-independent.

>I'm learning to like MAM, but the TieBreaker is a royal pain, since 
>it requires tracking individual ballots (or at least aggregates 
>thereof), not just the summary matrix.    Also, any method that 
>includes a random element is harder to justify in a public debate 
>(though committees might accept it).

Agreed. The other reason why I dislike the random element is that I 
believe it may mask a genuine tie between two or more alternatives. 
While such a tie would likely be resolved via a random selection, it 
would seem to make more sense to report the tie and allow the users 
of the method to determine how to resolve it.

>Under what circumstances would MAM-d give a worse winner than 
>regular MAM, or at least allow more strategic manipulation?

Unknown. I have not done a detailed analysis comparing the two methods.

>In regards to MAM, I was also wondering whether allowing 
>equal-rankings for non-bottom candidates (as opposed to strict 
>ordering) makes such Ties more or less likely.

Interesting question. However, I would guess that MAM would report 
fewer ties then MAM-d (assuming this is an appropriate name).

>Or is equal-ranking important for other strategic reasons, to avoid 
>requiring Favorite Betrayal?

I cannot answer this question either.