Social Media as Prediction Tool

Even as several social networks have surpassed the populations of most nations — more than 500 million people are now signed up for Facebook and 175 million for Twitter — we still tend to regard these sites in terms of their value to us as individual users. In the past year, however, social scientists have begun looking more broadly at the aggregate value of social media. According to a number of recent studies, it now seems possible that the networks’ millions of posts and status updates are adding up to something culturally and financially priceless.

This past April, for instance, Sitaram Asur and Bernardo Huberman at HP Labs demonstrated that by analyzing the positive or negative sentiments expressed in 2.8 million Twitter messages about 24 movies, they could predict how the films would perform at the box office. Their methodology — an algorithm, actually, that their company is now in the process of patenting — worked significantly better than the Hollywood Stock Exchange, another popular tool for predicting box-office success.

In October, a team led by Johan Bollen at Indiana University reported that by classifying 9.7 million Twitter posts as falling into one of six mood categories (happiness, kindness, alertness, sureness, vitality and calmness) they could predict changes in the Dow Jones Industrial Average. As Bollen explains, when he began his study, he expected that the mood on Twitter would be a reflection of up and down movements in the stock market. He never imagined it would be a precursor.

Twitter’s daily feed challenges some of the conventions of social science. Those old methods of fact-gathering — finding a small population for study, applying for a grant and then spending years doing observations or surveys — can seem quaint by comparison. Of course, it’s hard to ignore the mercenary aspects of big data, too. Any technology that might help predict the future does not come cheap. Twitter and Facebook — which last year used an analysis of posts to create a daily happiness index — can easily mine information about consumer products, in real time, from their data streams. “Given that there’s a population of 250 million people or more logging in every day, it seems ridiculous that you would try to get in touch with them through a phone,” says Cameron Marlow, the head of Facebook’s data-science team. “So it seems like it’s improbable that social media won’t be the way that we acquire opinion research.”

via