Ian Ayres has a fun guest post on the Freakonomics blog.  His argument is that in many decision-making situations it’s best to have some degree of randomness in your decision, so your opponent can’t just play for you to do the "obvious" best move.  (These ideas are very relevant to me, since one of my favorite squash opponents is extremely good at reading my next shot.  Especially when I have a very likely winning position, he’ll guess what I’m going to do next, run to the killer position for that shot, and end up stealing the point.  Even knowing that I need to surprise him hasn’t helped so far.  I need a random number generator in my racket!)

The post also shows another example of human decision-making flaws: football coaches don’t go for it on 4th down nearly as often as the data suggests they ought to.  (There are a few notable exceptions, including Bill Belichek of the Patriots.  Interestingly, announcers *still* call him out for being inappropriately aggressive on 4th down.  How many Super Bowls does he have to win?!)

One commenter points out a possible evolutionary argument: perhaps football coaches who are aggressive look stupid when they lose, and hence get weeded out.  Fun argument!  Notice, however, that it requires an
accepted wisdom to conform to.  Where did that come from?  Perhaps it’s
possible to motivate the accepted wisdom from known human decision-making flaws: we tend to value something we already have more than something we might obtain in the future, irrationally.  (For instance, most people would value $100 more than a 60% chance at$200.) So, the accepted wisdom for football coaches might be to take the points they have "in hand", because the administrators who fire them might think they’re stupid if they don’t — even if the expected value computation says "go for it!".

There’s a related, but different argument in "Wisdom of Crowds": the players might lose heart if they go for it and miss, which might reduce their performance in the rest of the game, and hence their *actual* expected number of points.  This sort of psychological effect might be amplified by the ordinary human preference for the bird in the hand described above.

I’m particularly interested in these areas in which human decision-making is unable to take advantage of what the data suggest.  Could a decision aid help?  What would a decision aid for a football coach look like, for instance?

John