Abstract

In a series of formal studies and less formal applications, Hong and Page offer a ‘diversity trumps ability’ result on the basis of a computational experiment accompanied by a mathematical theorem as explanatory background (Hong & Page 2004, 2009; Page 2007, 2011). “[W]e find that a random collection of agents drawn from a large set of limited-ability agents typically outperforms a collection of the very best agents from that same set” (2004, p. 16386). The result has been extremely influential as an epistemic justification for diversity policy initiatives. Here we show that the ‘diversity trumps ability’ result is tied to the particular random landscape used in Hong and Page’s simulation. We argue against interpreting results on that random landscape in terms of ‘ability’ or ‘expertise.’ These concepts are better modeled on smother and more realistic landscapes, but keeping other parameters the same those are landscapes on which it is groups of the best performing that do better. Smoother landscapes seem to vindicate both the concept and the value of expertise. Change in other parameters, however, also vindicates diversity. With an increase in the pool of available heuristics, diverse groups again do better. Group dynamics makes a difference as well; simultaneous ‘tournament’ deliberation in a group in place of the ‘relay’ deliberation in Hong and Page’s original model further emphasizes an advantage for diversity. ‘Tournament’ 2 dynamics particularly shows the advantage of mixed groups that include both experts and non-experts. As a whole, our modeling results suggest that relative to problem characteristics and conceptual resources, the wisdom of crowds and the wisdom of the few each have a place. We regard ours as a step toward attempting to calibrate their relative virtues in different modelled contexts of intellectual exploration.