[EM] Bouricius reply, BR, Tideman, recent USA elections

[robert bristow-johnson]

On Nov 9, 2009, at 1:46 AM, Jonathan Lundell wrote:

> On Nov 8, 2009, at 10:40 PM, robert bristow-johnson wrote:
>
>> On Nov 8, 2009, at 7:50 PM, Jonathan Lundell wrote:
>>
>>> On Nov 8, 2009, at 4:35 PM, Warren Smith wrote:
>>>
>>>> Tideman said IRV was unsupportable if it is feasible to compute
>>>> pairwise matrix.  That was
>>>> because Tideman had other voting methods he considered clearly
>>>> superior to IRV and these methods used the pairwise matrix.    By
>>>> "clearly superior" I mean, so superior in every respect, that  
>>>> Tideman
>>>> felt there was no conceivable use for IRV, ever (in situations  
>>>> where
>>>> it was feasible to compute pariwise matrix) where that use could be
>>>> "supported."
>>>> That is what "unsupportable" means.
>>>
>>> Tideman ranks IRV highest in resistance to strategy, and  
>>> generally better than the pairwise methods in lucidity
>>
>> can someone explain to this layman what the metric "lucidity" is  
>> in regard to election methods?
>
> Generally speaking, the degree to which a method is understandable  
> by voters.

then i think that "the pairwise methods" (i presume Condorcet) is  
getting a bad rap, particularly compared to IRV.  Condorcet is simple  
conceptually (save for what to do in case of a cycle).  dealing with  
a cycle can be a mess, but if there is no cycle, executing the  
Condorcet tally is a piece of cake.  any C programmer can punch that  
program out in a half hour.

and i think that tactical or strategic voting is most commonly a  
result of voter regret.  usually the tactic is "compromising" because  
some Nader voters in Florida figgered out that in 2000 they just  
helped put the worst and most illegitimate person in U.S. history  
into the Oval Office and decided to vote for Kerry (even though they  
don't like him) the next time around (2004).  "compromising" is the  
tactical insincere vote that is most prevalent, in my opinion.  i've  
seen first-hand how IRV will make some Republicans think twice  
(assuming IRV survives the referendum) about voting for their  
candidate as their first pick, because they found out that doing so  
actually elected their least favorite candidate (the Prog) over the  
more centrist candidate (the Dem, whom more would prefer over the  
Prog).  that would not have happened with Condorcet in that example.   
i really don't see IRV as more resistant to tactical voting in regard  
to the most common tactic of compromising.

>>
>>> and cost of computation.
>>
>> and why the cost of computation (as if it takes the official  
>> computers 10 seconds to crunch the numbers instead of 5) is  
>> important in modern times?  it's not like the cost is O(2^N) or or  
>> O(N!).
>
> It's relevant in two ways, I think. One is wrt hand counting.

it increases the cost of hand counting by a factor of M*(M-1)/2 if M  
is the number of candidates.

> The other is actual computational complexity. I don't think that  
> any of the methods on Tideman's FPTP-replacement list are complex  
> in that sense, but there are certainly proposed systems (including  
> Tideman's own elaboration of PR-STV) that appear to be impractical  
> to compute with current technology and algorithms, in some cases.

the number of pairs in Condorcet is M*(M-1)/2 for M candidates.  your  
computer runs M*(M-1)/2 hypothetical traditional elections for each  
pair.  once the physical ballots are recorded, it's a line of data in  
a file (the optical scan or punch cards need not be processed  
again).  if there are N voters, the cost is N*M*(M-1)/2.  it's  
proportional to the big number (the number of voters).  the number of  
candidates is small (is M ever bigger than 6 or 8?)  i cannot see the  
computational cost of Condorcet as anything approaching prohibitive,  
even for a national election with a dozen candidates.

--

r b-j                  rbj at audioimagination.com

"Imagination is more important than knowledge."