Looking at a half-million tweets over 15 months, a first-of-its-kind study from Georgia Tech has revealed a set of reliable predictors for building a Twitter following.

The research was performed by Eric Gilbert, assistant professor in Georgia Tech’s School of Interactive Computing. Gilbert found that Twitter users can grow their followers by such tactics as:

• Don’t talk about yourself.Informational content attracts followers at a rate 30 times higher than content focused on the tweeter. The study found users talked about themselves in 41 percent of their tweets on average.

• Be happy.Twitter is mainly based on weak social ties (most followers do not know each other offline), which makes it more important to stay away from negative posts such as death, unemployment and poor health.

• Cool it on the hashtags. While hashtags are definitely useful tools for expressing emotional commentary or tying tweets to larger events or issues, they can be abused. Researchers found that the higher a Twitter users’ “hashtag ratio,” the less likely they were to attract new followers.

“To our knowledge, this is the first longitudinal study of follow predictors on Twitter,” Gilbert said. “For the first time, we were able to explore the relative effects of social behavior, message content and network structure and show which of these factors has more influence on the number of Twitter followers.”

Working with Ph.D. student C.J. Hutto and Sarita Yardi, now an assistant professor in the University of Michigan’s School of Information, Gilbert examined the tweets of more than 500 Twitter users. After identifying 2,800 terms that convey positive and negative emotions, the team scored each term based on a sliding scale of positivity. They were then able to determine whether Twitter users who used each term gained or lost followers.

The team discovered that certain identifiable strategies in message content and interaction with other Twitter users, as well as the structure of one’s Twitter network, have a predictable effect on the number of followers. For example, Twitter “informers” (users who share informational content) consistently attract more followers than “meformers” (users who share information about themselves).

“Followers are Twitter’s most basic currency, yet little is understood about how to grow such an audience,” said Gilbert. “By examining multiple factors that affect tie formation and dissolution over time on Twitter, we’ve discovered information that could help technologists design and build tools that help users grow their audiences.”

The team’s findings are summarized in the paper, “A Longitudinal Study of Follow Predictors on Twitter,” which will be presented in early May at the ACM SIGCHI Conference on Human Factors in Computing Systems in Paris, France.

To view research by other Georgia Tech researchers at SIGCHI, visit http://chi.gatech.edu.

Source: College of Computing at Georgia Tech