Humans versus Computers in Decision Making

I was listening to the "Mike & Mike" on ESPN radio this morning on my morning commute to work.  They were debating whether the method that the NCAA selects the final four teams for the college football play-offs is better than the old way (Bowl Championship Series, or BCS as it is more commonly known) or even if there was a better alternative.  As I understand it, the BCS used a rather complex computer algorithm that was based on the average of the AP and ESPN/USA Today coaches poll (i.e., a very subjective, human poll), "strength of schedule" (calculated by the win/loss record of opponents and cumulative wins/losses of a team's opponents' opponents), and margin of victory.  There was a lot of controversy about the BCS, but I think for the most part that it worked fairly well during the time it was in place (1998 through 2013).


Mike Greenberg and Mike Golic both had positive and negative things to say about the current system, which relies heavily upon subjective decision making by college football experts.  Interestingly enough, they seemed to prefer the BCS method of selecting the top four teams in college football (as opposed to their selection by a panel of experts), who would then play each other in the same play-off system that is currently in place to determine the NCAA National Champion. In other words, they both seemed to think that computer algorithms are preferable to humans! 


There is actually a lot of research that has been performed in this area.  Dr. Paul Meehl, a Professor of Psychology actually published a series of studies that explored this topic in a variety of settings, including the clinical setting!  His 1954 book, Clinical and Statistical Prediction: A Theoretical Analysis and a Review of the Evidence suggested that formalized statistical algorithms and decision rules actually performed better in a variety of settings compared to informal, subjective, "gut instinct" types of decision making.'


I am not saying that "gut instinct" is never right.  On the contrary, there are some good studies that suggest that "gut instinct" can be useful (see the study by Beth Crandall for a really good example - we have used this method to improve situation awarenessin our hospital).  However, the evidence to suggest that use of objective decision rules and algorithms can be a powerful tool is very compelling.