It is vital to consider the question of error covariance, which can be described as the tendency to make similar mistakes. Please see the error covariance website for more information.
An interesting approach to dealing with this problem is to match people with others who have demonstrably low error covariance. If you had ten people to work on estimation problems, matching them into five teams of two in such a way as to minimize the overall error covariance. This can be done with weighted bipartite matching, where the weight of a potential link was greater if the error covariance was lower.
A possible application of this is in the justice system, in which carefully matching judges could reduce the total number of wrong decisions in the whole system.