[EM] Dave: Condorcet

[Kristofer Munsterhjelm]

On 04/02/2012 03:49 PM, Abd ul-Rahman Lomax wrote:
> At 08:57 PM 4/1/2012, robert bristow-johnson wrote:

>>
>> Approval applies a burden of tactical voting to the voter right from
>> the start.
>
> In other words, a burden of responsibility for the effect of our actions
> on the world. Those who would protect voters from taking responsibility
> for their actions are those who really don't care about democracy, which
> requires this responsibility. Note that Approval is simply Plurality
> with an option. With an option comes an increase in power, with an
> increase in power comes an increase in responsibility.

Keep in mind that voting in itself is a way of easing a burden. Ideally, 
we'd all get in one big town-hall meeting and keep discussing until 
there's a majority - or supermajority, if you like consensus. However, 
that's impractical. So instead, when there are too many people, we vote.

Approval requires data so that the voters can make the decision about 
where to put his Approval threshold. In that respect, it requires more 
coordination than the (good) ranked voting systems. If you're willing to 
accept that need for coordination so that you can get better results 
with a simpler system, then that's alright, but others might disagree.

I don't think it's inherently contradictory for Robert to want to just 
have to rank without having to take into account "his responses to their 
responses to his responses" through rounds of polling. He simply wants 
to be closer to the one-off voting end of the axis, farther from 
town-hall interaction end.

> Much of the difficulty here comes from adopting an anti-democratic
> principle, election without a true majority approval. Basic democratic
> process requires making no decision without majority approval. In the
> name of "efficiency" or "cost savings" -- same thing -- we give up
> democracy. That's an old story, eh?
>
> Within standard process, where elections fail (and presumably are
> repeated, then), if there is no majority approval for an outcome,
> Approval can be seen as a simple and fair way to make the process more
> efficient. Don't know whether or not to Approve So-So? Don't. But that
> might cause the election to fail! We'll need to keep voting or the
> Office of Dogcatcher will go vacant! Sure. Don't want that cost, approve
> So-so. There is still no such thing as a free lunch.

In a practical modification of Plurality, as is discussed here, not 
approving So-So wouldn't trigger an additional round to home in on 
consensus. Instead, Worse might win with less than majority approval. 
The convergence wouldn't be built into the system, but rather would have 
to be done by the means of polling, and I think that's quite fragile.

>> perhaps someone will want to bury a candidate they sorta like but who
>> is not their favorite, to help their favorite win. if they don't
>> Approve, then how much Bayesian regret will result when they find out
>> their favorite was not in the running at all and their 2nd choice lost
>> narrowly to someone they hated? if they do Approve, how much regret
>> will result when their favorite lost narrowly to their 2nd choice?
>
> Oh, my head hurts! You mean I actually have to think and weigh options?
> They didn't warn me about this when I registered to vote!

"Oh, my head hurts! You mean I actually have to get into the meeting and 
try to convince a majority of the thousands already there? They didn't 
warn me about this when I registered to vote!"

Granted, that's a bit unfair, but I'm merely trying to show that one can 
reasonably prefer a method that requires less external convergence to 
one that requires more. Perhaps the results will be worse (perhaps not). 
But the burden will be lighter, too.

> If you don't trust your Favorite to make further decisions, why in the
> world would you want to elect him or her to the job?

That's a good statement for parliamentarism, as well. Though I prefer 
parliamentarism, let's consider why someone might prefer direct 
elections to Asset-like systems. I think it involves quantization error.

Simply put, a majority of a majority is not the same thing as a 
majority. Similarly, the favorite of your favorite may not be the same 
as your favorite. So it's not that cut-and-dried.