Wisdom of Crowds

  Scott E Page  - Coursera  The Diversity Prediction Theorem

Where does collective wisdom come from?

People make predictions based on different models.   We know more accurate individuals lead to more accurate predictions. 

However, in addition, a more diverse crowd leads to more accuracy.

Example :

Amy predicts 10   error (10-18)^2  = 64

Belle predicts 16   error  (16-18)^2 = 4

Carlos predicts 25 error (25-18)^2 = 49     

Average error = Total error/3     = 117/3   = 39

Average value is 17

Actual value is 18

How accurate is the crowd?  Crowd error  =  (17-18)^2 = 1

Diversity

Amy   (10-17)^2 = 49

Belle  (16-17)^2 = 1

Carlos (25-17)^2 = 64

Average diversity = 114/3 = 38

 Because 

In the example the crowd error 1= the average error 39 –  the diversity 38

For the wisdom of crowds to occur the crowd error needs to be small or it is not wisdom.  The average error needs to be large or it is not a hard problem, so the diversity must also be large.   The diversity comes from people using different models.   Individual ability and diversity are equal partners.

The Madness of Crowds comes from like-minded people that are all wrong.

Large CE  + Large AE – Small diversity