One question of interest in sociology and journalism is what makes some piece of text objective. I believe this question can be answered using machine learning. I trained a classifier to distinguish between movie synopses (which describe what happens in a movie) and movie reviews (which describe a person's opinion about a movie). It was able to distinguish not only between synopses and reviews, but between opinion pieces and other news reporting in The New York Times. Moreover, the objectivity ratings correlated moderately with ratings of the same articles made by a sample of undergraduates.

Check out a video of my talk on this topic.

Another topic of interest in the social sciences is the wisdom of the crowds. If people's predictions are taken in aggregate, they are surprisingly good at estimating unknown quantites (like the weight of an ox) and predicting future events. Recently, researchers have been trying to select "super-forecasters" for a group that would make predictions better than an unscreened crowd. My research suggests that text information provided by these people may be a rich source of information. In particular, I showed that people who sound more positive in their messages tend to make better predictions.

Check out a poster about this project.