[EM] language/framing quibble

[Kristofer Munsterhjelm]

Raph Frank wrote:
> One option is to select the legislature at random.  Stratified random
> sampling would yield a highly representative legislature.  The
> population would be split into N groups, such that each group is
> reasonably homogeneous and then 1 person picked from each group.  This
> also reduces the benefit from corrupting the random process.  Also,
> corrupting the stratification just increases the random variance, it
> doesn't actually change the expect result.  Corrupting both means that
> you get to pick the legislature.
> 
> This has the advantage that it eliminates the point in campaigning.
> Every 5 years, a group of people get a mail in the post informing them
> that they have been selected for 'legislature duty' .. though unlike
> Juries they would presumably be paid.
> 
> The disadvantage (or advantage depending on your viewpoint) is that it
> leads to a legislature made up of average people.
> 
> I have suggested that a way around it is to have a multi-stage
> process.  The people picked at random are asked to select the 'person
> they know who they would most respect to hold office' and that
> generates a second group.  The rule would require that the person
> picked is somehow connected to them, say friends or family members.
> After a few stages, say 10, the final group becomes the legislature.

If duplicate votes don't count, then there'll be a natural incentive to 
pick friends instead of central party figures. All campaigning would do 
would be to give whichever candidate's being promoted a lot of votes, 
which is no better than the candidate in question getting a single vote.

The process would have to run multiple rounds if different voters vote 
for the same person, since that would exhaust the legislature. One way 
of doing this would be to send mail to some number of people. Then 
iterate 10 times (in the way you specify) until you have a set that'll 
be on the council. If multiple voters voted for the same person, then 
the size of that set will be smaller than the council, so send out more 
mail (to yet other people) while keeping the current set secret. Repeat, 
getting a new set, until the union of the sets is large enough. If there 
are too many, just do a random sample within the union of the sets.

There's still some temptation to try to become famous, since if you're 
famous, a greater subset may pick you. However, there voters have an 
incentive to make their vote count by not picking someone who's too 
famous (since somebody else would have, already). In the worst case that 
some want famous people on the council yet nobody votes for him, there 
would be instability, and the voters may reason that it's better to be 
sure and pick a friend instead, than have to second- (and third-, and 
fourth-...)guess what the other voters are going to do.

> This should result in a reasonably competent legislature (assuming
> each person picks someone more competent than themselves) and the rule
> that you must pick a friend/family members for each link means that
> campaigning is pointless.

I wonder if it could lead to extremism, where voters transfer their 
votes to friends that have "strong opinions". It might, but then again, 
it might not. Like delegate cascade, the dynamics aren't obvious.