[EM] Arrow's axioms & an alternative to elections

[Ernest Prabhakar]

Hi Philippe,

On Mar 5, 2004, at 11:03 PM, Philippe Errembault wrote:

> My point is that if you want to rank multi-dimensional information, 
> you will have to project your space to a one-dimensional space. This 
> will be done using a function that, especially for human beings, will 
> depends on your mood, on context, on... I do not know what else... the 
> fact is that this will not be stable data.

That's partly true, but I would argue that's a fundamental part of 
being human: the tension between our rationality and our emotionality.  
  In fact, I think that's one of the ways we mature, by learning to 
align our transient moods with our integral convictions and principles. 
   To try to avoid that seems to me like an abdication of 
responsibility.

> Even more: the space of this information could even be non Euclidian, 
> because it's probably easy to determine preferences between two 
> things, but if you try to establish the graph of all your
> preferences, I'm not sure you will easily obtain an acyclic graph. 
> Just try to rank all your friends on the basis: which you prefer 
> most... not quite easy, isn't it !?

That's where I think I disagree with your model. Elections are not an 
abstract ranking.  They are a specific question.   For example, if the 
question is "Which friend would I want to stand by me in a dark alley?" 
or "Who do I think should organize our next party," its pretty easy to 
come up with a non-cyclic ranking.  I think elections are more like 
those questions, than an arbitrary ranking of friends.

> So, what I suggest, is to have people chose one, let's say: "level 
> one" representative, which will eventually be themselves, the 
> obligation they have is to give this representative, proxy to vote for 
> them. "Level one" representatives and only them will have to vote (and 
> be obliged to) when there is something to vote for. Let us put 
> restrictions on this: 1/It is strictly forbidden to chose as 
> representative, someone that could have a power on you. 2/ a 
> representative can only represent, let us say, at most (e.g.) 19 
> persons. To be a representative, you at least need to represent 
> yourself.
>
> One of the most important advantages of such a process, is that having 
> more intermediate levels, gives more hindsight (not sure about the use 
> of this word...) to the decision process, which could make it smarter. 
> We could also compare this principle to a multi layer neural network.

So, I think there's an interesting aspect of the value of hindsight.   
But, this idea of indirect democracy seems pretty close to what we in 
the U.S. used to have for an electoral college; its not much different 
from having state legislators elect federal representatives.  In 
practice, the actual effect seemed to be actually to *decouple* the 
interests of representatives from the low-level electorate.

In the U.S., I would argue the purpose of democracy is not so much to 
choose the best leader (since that's not well-defined), but to hold 
leaders accountable for their actions.  Direct elections at least have 
the advantage of motivating leaders to at least appear to be serving 
the needs of their constituency (the whole population).  It may be a 
small advantage, but I suspect it is a crucial one.

Direct elections further require the electorate to be the ultimate 
arbiters of whether leaders are acting in the population's best 
interest.   One of the tenets of American democracy is that this is a 
fundamental right -- and a fundamental obligation -- that nobody should 
or can delegate.  Even if people choose to follow someone else (e.g., a 
party leader) I believe democracy is healthiest when such delegation is 
*not* managed or enforced by the state.   I think things like 
rank-order voting are a better solution to the split-vote/runoff 
problems like what you describe in France.

Of course, maybe that's just my American perspective, but I think it is 
at least a significant philosophical bias, not just ethno-centrism.

-- Ernie P.