Peer effects at the Air Force Academy

I heard this story the other day on NPR and thought it was absolutely fascinating.  Officials at the U.S. Air Force Academy wanted to see if there was a way to improve the academic performance among lagging students, so they were paired with the best performing students to see if peer effects would boost grades.

Alas, their grades got worse.  It turns out that, even though they were supposed to work together, the two classes of students split off and the poor students suffered.  It seems that the average-performing students act as a sort of "glue" when working together as a team.

But in a further twist, when the middle students were isolated from the effects of the high and low-performing students, they improved academically:

Well, the middle group started to do better, Linda. When they didn't have the influence of either the better students or the worst students, the middle students by themselves, their test scores went up. And for the policymaker, this presents a dilemma, because it suggests that in regular squadrons where everyone is mixed together, the performance of the middle students is actually adversely affected. You pull them out, they start to do better. But when you pull out, the weakest students start to do worse.

So it's difficult to find the right balance, it appears.