How does collective intelligence outperform experts?

Question

I am interested in doing several collective intelligent programs, but I wonder how this can work?

They say that he can give accurate forecasts: for example, the book of collective intelligence "O'Reilly" says that a collection of actions by traders can really predict future prices (for example, corn) better than an expert can be.

Now we also know that in the statistics class, if this is a room of 40 students who pass the exam, there will be from 3 to 5 students who will receive an “A” grade. There may be 8 who get a “B,” and 17 who get a “C,” and so on. That is, basically, the bell curve.

So, from these two points of view, how can the collection of answers “B” and “C” give a better prediction than the answer that received “A”?

Note that the price of corn, for example, represents the exact price factor in the weather, the demand for food companies using corn, etc., rather than a “self-fulfilling prophecy” (more people buy corn futures and prices rise, and more people buy futures again). It actually predicts supply and demand accurately to give an accurate price in the future.

How is this possible?

Update: can it be said that collective intelligence will not work on stock market euphoria and panic?

Answer 1

Wisdom the Crowds .

, . .

Answer 2

, "" :

guess = right answer + error

, . - error ( ), .

, - ( B C, A), B C, .

, , , ...

Answer 3

, , , (, ) . - , , . , . , , , .

, . , , - , , , . , , . , , . , , , , , .

, , . . .

Answer 4

, , / . .

Answer 5

, . ( , ), .

IOW , ( ).

, . , , , , . , , . ( )

The problem with the class analogy is that the classification system does not imply that students are masters in their (difficult to predict) locality, so it is not comparable.

Ps note that the basic axiom depends on the fact that all players are experts in a small part of the field. You can discuss whether this requirement really tolerates a website environment.