Twitter Drives New Insights in the Social Sciences

Twitter, Facebook and other online social networks are letting scientists take their research out of the lab and into the real world. Their research could radically change the way we view ourselves and the people around us.

In the past, scientists studying the web of relationships between individual humans and how they affect behavior had to choose between two tools: surveys, which often downplay the impact of a social network on decision making, and direct observation, which necessarily involves fewer participants than most real-world social networks. But online social networks have added a powerful new tool to the arsenal: huge communities of people whose interactions can be recorded and analyzed, providing an unprecedented look at the role of social networks in shaping who we are. Recent studies are revealing new insights about human behavior in ways that weren’t previously possible.

“Social scientists have not really been able to study social networks, except on a very limited scale,” said Michael Macy, a professor of sociology at Cornell University. “Now we’re able to study social networks on a population-level scale” thanks to Twitter.

One of the biggest benefits of digital networks is their ability to help scientists account for homophily, the fact that people adopt the views and preference of the people they’re closest to. Homophily has posed the social sciences’ ultimate chicken-and-egg question: Do your actions influence the actions of your friends, or are your friends doing the things you do because people with similar interests and habits tend to flock together?

Why Do You Vote?

Take, for instance, a landmark 2009 study by Nicholas A. Christakis of Harvard and James H. Fowler of the University of California, San Diego. The results suggested that a person’s decision to vote can influence hundreds of people linked through their social network to head to the polls. 

While highly regarded by social scientists, that study, as well as similar research by the same researchers concluding that obese people influence their friends to put on weight, faced the homophily question. Do your friends vote because you vote, or do people who have an interest in politics tend to associate with one another? Did your obese friends cause you to become overweight, or did you choose obese friends because they’re similar to you?

Fowler has been using Facebook to try to answer those questions, according to Macy. With Facebook, researchers can track pre-existing relationships as well as the various interests of network members to gauge the effect of homophily, if any. While the work is preliminary, Macy is encouraged to see that previously unanswerable questions might now be resolved. 

“In the past, we could either do small-group, observational studies, but the groups tended to be very small, or we could do surveys and random samples, but the data did not really reflect the influence of the social network,” Macy said. “Online data from social networks marks the first time we’ve been able to look at these questions on such a large scale.”

Understanding Why Dogs Don’t Bark

Studies based on data from Twitter, Facebook and other social networks have proliferated over the past five years, Macy said. In addition to tracking how concepts, viewpoints and ideas spread through social networks, social scientists have, for the first time, been able to look at the concepts that don’t spread through social networks.

“When you think about it, understanding why things don’t spread is really important,” the researcher said. “We’ve never been able to really look at why some dogs don’t bark, and why some information has a longevity and persistence.” 

One of Macy’s students, for example, is working on research that suggests that information that goes viral on a social network “tends to be a flash in the pan.” Information that spreads more slowly, getting discovered and rediscovered over and over again, has more staying power.

How Happy Are You? It Depends On When You’re Reading This

In his own study, Macy analyzed words expressing mood in 500 million tweets sent between February 2008 and January 2010. He found that people’s mood, as measured by their tweets, tended to be elevated in the morning and to decline as the day progressed. Weekends tended to be “happier” days, although mood peak started later in the morning, possibly reflecting twitterers’ tendency to sleep in. Most remarkable, Macy said in an interview with Science magazine, was that the tweets showed a similar pattern across the 84 countries where the tweets originated, suggesting that the daily mood curve is fundamentally human rather than cultural. In the past, such a study would have been limited to a smaller population within a lab, meaning researchers couldn’t empirically compare similarities and differences in mood across cultures. 

A larger study by University of Vermont scientists led by sociologist Peter Dodds analyzed mood-related words in 4.6 billion tweets sent over three years. The results mirrored Macy’s. However, the Vermont team found a pattern of extraordinary, or “outlier,” days. People were markedly more happy on holidays and other special occasions, while unexpected negative events, like the 2011 earthquake and tsunami in Japan, led to a global downswing.

The Vermont study also identified a troubling trend: Global mood headed downward between April 2009 and the first six months of 2011. Ultimately, though, such results could lead to better public policy, Dodds said. While tracking tweets “sounds a bit Orwellian,” he told Science, policy makers could pair a mood index with economic indicators to better understand economic trends.

There are more studies than we’d ever have time to discuss, and many of them are quite exciting,” Macy said. “We’re going to be able to better understand what role social networks play in shaping our lives better than we ever have before.”

Facebook Comments