A little randomness can go a long way

The following excerpt is from Professor Bruce Sacerdote ‘s article "What to expect when a college assigns students to random roommates".

"Duke University recently announced that first-year students will now be randomly assigned to their dormmates. The goal is to give students a chance to meet and learn from peers from a completely different background. Is this silly social engineering or smart policy?

First, let’s acknowledge that roommates and dormmates matter for students’ social networks. In one study – titled “How Do Friendships Form?” – David Marmaros and I examined email behavior among thousands of college students. First-year roommates exchange 45 times more emails with each other than with a randomly chosen member of the incoming class.


The bonds that students form their first year are long lasting. By senior year, former first year roommates are still emailing each other at 10 times the rate that they would with other students on campus.


What are we to make of all this? The bottom line is that as human beings we naturally gravitate towards our comfort zone and find peers who look a lot like ourselves, as shown in the email study and many other studies of friendship groups. But our natural instincts do not always benefit us in the long run.

College provides students an opportunity to meet different people, expand their boundaries and think about majors, careers and ideas that were not part of their previous experience. In short, in a world where people choose to interact with others from a similar background, a little randomness can go a long way."

See other research outcomes referenced in the articles : "How Do Friendships Form?", "Peer Effects with Random Assignment: Results for Dartmouth Roommates"