Twitter and Bin Laden 2

The burst of Twitter reports of Bin Laden’s death prior to any official announcements is continuing to generate interest.  NPR has an informative interview with Andy Carvin, NPR’s senior strategist for social media (thanks to Georgia Kernell for sending it to me).  The event raises several interesting questions about social dynamics and the spread of information.  It would be interesting to know how many false reports of Osama Bin Laden’s death there have been on Twitter prior to this that didn’t “go viral.”  It would seem that a key feature of this particular cascade of tweets was the apparent authority of the “earlier adopters.”  The initial tweet that got things going seems to have come from a former aide to Donald Rumsfeld.  This tweet was probably picked up by other “Beltway insiders.”  I would guess a key feature to the rumor’s success in spreading was the fact that these initial tweeters were people that other people trusted to know this sort of thing.  That’s a difficult aspect to capture in basic compartmental models of rumor spreading, or even standard network models of information transmission.  You not only need to have highly credible individuals, but you have to have a large cluster of credible individuals that are all tightly connected to one another.


Robert Shiller: We Need Better Crystal Balls

In today’s New York Times, Robert Shiller argues that the recent crisis should precipitate a movement to collect better data that could help in predicting similar events in the future.  He likens the problem to predicting hurricanes; something that was impossible at one time, but which we can now do accurately because of better data.  Shiller’s point is a sharp contrast to the perspective of Duncan Watts espoused in his recent book, Everything is Obvious (Once You know the Answer). Watts argues that no matter how good the data, predicting many things is just impossible, in part because its impossible to even know what it is you should be predicting.  Shiller’s opinion on this matter has an extra sense of authority of course, because he is one of the few economists who saw the housing bubble/credit crunch/financial crisis coming and said something about it.  He also forecasted the stock market collapse of 2001.

So, who’s right?  Is Shiller just another example of luck and large numbers?  After all, a lot of people are making a lot of predictions out there, so we shouldn’t be surprised that someone got it right.  And Watts’ argument that in many cases we can’t know what we should have been predicting until we put the outcome in a context has merit.  Both are right to an extent.  I do believe the crisis was predictable, and that Shiller wasn’t just lucky in foreseeing it, but I also agree with Watts that more data isn’t the full solution.  We also need to know where to look — how to put that data together.  We can forecast hurricanes today not just because we have better data, but because we have better models and a better understanding of how the weather works.  It’s a little surprising that Shiller doesn’t point this out, because he is also well-known for arguing that economists should in part be looking in a different direction than they have in the past.  Namely, our economic models need to incorporate more social and psychological effects (see his recent book with George Akerlof, Animal Spirits).

Content Analysis for Predicting Future NFL Player Success

As anyone with a fantasy football team can tell you, predicting the success of NFL players in their rookie year is a tough business.  Now NFL teams may have a new tool: content analysis.  A company called Achievement Metrics is analyzing the content of college player’s post game press interviews to predict characteristics like leadership, discipline, and behavior risk.  They use the same techniques to predict how likely an individual is to be a terrorist.    There’s an article on their approach here.