In recent  years, the predictive element of social science has been discussed by a few people: how much can we use data from the past to predict the future? In an interview with Scientific American, a mathematical sociologist who works at Yahoo! Labs talks about our predictive abilities:

A big part of your book deals with the problem of ignoring failures—a selective reading of the past to draw erroneous conclusions, which reminds me of the old story about the skeptic who hears about sailors who survived a shipwreck supposedly because they’d prayed to the gods. The skeptic asked, “What about the people who prayed and perished?”
Right—if you look at successful companies or shipwrecked people, you don’t see the ones who didn’t make it. It’s what sociologists call “selection on the dependent variable,” or what in finance is called survivorship bias. If we collected all the data instead of just some of it, we could learn more from the past than we do. It’s also like Isaiah Berlin’s distinction between hedgehogs and foxes. The famous people in history were hedgehogs, because when those people win they win big, but there are lots of failed hedgehogs out there.

Other scholars have pointed out that ignoring this hidden history of failures can lead us to take bigger risks than we might had we seen the full distribution of past outcomes. What other problems do you see with our excessive focus on the successful end of the distribution?
It causes us to misattribute the causes of success and failure: by ignoring all the nonevents and focusing only on the things that succeed, we don’t just convince ourselves that things are more predictable than they are; we also conclude that these people deserved to succeed—they had to do something right, otherwise why were they successful? The answer is random chance, but that would cause us to look at them in a different light, and changes the nature of reward and punishment.

Interesting material and Watts’ just published book (Everything Is Obvious: *Once You Know the Answer)  sounds worthwhile. There are also some interesting thoughts later in the interview about how information in digital social networks doesn’t really get passed along through influential people.

I haven’t seen too much discussion within sociology about predictive abilities: how much do we suffer from these blind spots that Watts and others point out?

(As a reminder, Nassim Taleb, in his book Black Swan, has also written well on this subject.)