Skewed Populations and Sarcasm Among Human Subtleties Mustafaraj Points Out

August 2, 2012

USA Today spoke with Eni Mustafaraj, Norma Wilentz Hess Fellow in Computer Science, for a story on the new Twitter Political Index, also known as the “Twindex.” According to Twitter, the tool is a “daily measurement that evaluates the sentiment of Twitter users’ feelings about the candidates, Barack Obama and Mitt Romney, relative to the more than 400 million tweets sent about all other topics each day.”

Mustafaraj explained some of the problems in relying on such a tool to accurately measure the moods of voters.

“Without human intelligence, analyzing the sentiment of 140 characters can be very difficult. For example, many tweets are sarcastic, but computers cannot understand contextual sarcasm,” she said. Read the article, "Twitter index tracks sentiment on Obama, Romney" in USA Today.

Mustafaraj pointed to the prevalence of humor and play on Twitter, such as the popularity of hashtag games. Such games can pose an interpretation challenge to tools like Twindex.

Mustafaraj also cited the ability of sentiment tools to influence behavior on social media. “Once people know that they are being observed, and their tweets are being used for a certain purpose, they tend to change their behavior to game the system,” she said.

Furthermore, Mustafaraj said that despite the high volume of messages, the Twitterverse is “a very skewed sample of the population.” A recent study she co-authored found that a small group of vocal Twitter users is responsible for generating the largest volume of tweets. Co-written with Panagiotis Metaxas, professor of Computer Science and founder of Wellesley’s Media Arts & Sciences program, and students Samantha Finn ’12 and Carolyn Whitlock ’12, the findings are published in the paper, "Vocal Minority vs. Silent Majority."

Working with Professor Metaxas, Mustafaraj has conducted extensive research on reliability of social media messages and the impact of those messages on perception, politics, and behavior.

In 2011 Metataxas and Mustafaraj were awarded a $492,000 grant from the National Science Foundation (NSF) to develop an application that examines the origin, authenticity, and trustworthiness of messages disseminated on social networks. The application will help users make a better determination about whether information can be trusted.

Wellesley students continue to make contributions to the research. This summer, Marie Vasek ’12, Diana Gerr ’15, and Stephanie Lee ’14 studied the spread of political rumors on Twitter, and designed tools to follow these rumors and the communities that play political hashtag games. They will present their research findings at Wellesley’s annual Summer Research Poster Session, to be held Thursday, August 2, in the Science Center.