Fascinating post. I would highly suggest that whatever models and parameters you use to simulate a "fractal democracy" should also be applied to other forms of democracy. For example, the knowledge one has about a system as a whole is inversely proportional to the size of the system.
You referenced a claim that markets are "fractal democracy"; however, markets are not person-weighted but wealth-weighted and wealth is Pareto-distributed and often not based on merit but on political connections. Markets help people reach a consensus on the allocation of private property through voluntary exchange. What markets do not allocate is political or social power and often political and social power can corrupt the markets.
So any simulation must assume the following there is some widely held opinion of the people to be discovered and then ascertain the accuracy and precision of the measurement of that widely held opinion.
If you utilize end-justifies-the-means thinking to judge fractal democracy then the analysis will be corrupted by the value system of the experimenter. So, merely assuming an "ideal best interest of the fractal" as a goal is missing the point.
Another factor missing from the analysis is a community being selective over its membership and actively excluding those who are contributing to the noise.
In my book More Equal Animals - The Subtle Art of True Democracy I outline a number of these points including:
Therefore, the purpose of meetings is to tap the wisdom of the crowd by having community members make local judgments about just the people in the group instead of asking them to make global judgments.
So if you were to simulate this you would want to ask yourself "what percent of available information" was evaluated and by how many people.
RE: Towards a statistical simulation of fractal democracy