SciTech

Twitter study looks for most effective ways to share info

Postdoctoral researcher Paul André led a study that aimed to understand how various types of Twitter updates were perceived by users. (credit: Jennifer Coloma/Staff) Postdoctoral researcher Paul André led a study that aimed to understand how various types of Twitter updates were perceived by users. (credit: Jennifer Coloma/Staff)

A 2011 CNN report indicated that there were over 300 million Twitter users. Despite these large numbers, users may not know how their followers react to their posts. There might be the occasional “unfollowing” or “retweeting,” but no additional feedback is typically given.

Presented with this problem, researchers at Carnegie Mellon University, the Massachusetts Institute of Technology (MIT), and the Georgia Institute of Technology recently published a study designed to understand the different reactions spurred by Twitter updates.

The study was led by Paul André, a postdoctoral fellow in the Human Computer Interaction Institute, and accompanied by MIT Ph.D. student Michael Bernstein and Georgia Tech Ph.D. student Kurt Luther. They will present their findings on Monday at the Association for Computing Machinery’s Conference on Computer Supported Cooperative Work in Seattle, Washington.

André said that the question that ignited this research was the following: There are many mechanisms of positive reinforcement, such as favoring tweets, “like” buttons on Facebook, or “+1” on Google, but what would a “dislike” button tell us? There is an accepted practice of online social interaction, as Carnegie Mellon computer science Ph.D. student Rob Simmons confirmed, but the norms are not clear to everyone.

“If everyone understood the best way to phrase something ... people could improve the way they write content,” André said.

The research aimed to raise awareness among Twitter users of what makes their tweets more likeable. The design implications are also valuable: A reader might be able to filter the Twitter feed content according to their preferences.

The researchers designed a website called “Who Gives a Tweet?” (WGAT), where followers could rate the tweets of other users. Users could provide feedback for a feed by clicking one of several options: “Worth Reading” (WR), “OK,” and “Not Worth Reading” (NotWR).

Users could also explain why they chose a particular rating by selecting one of the following: “funny,” “exciting,” “useful,” “informative,” or “arrogant,” “boring,” “depressing,” and “mean.” After WGAT launched, it was featured on news sites like CNN, Mashable, TechCrunch, and OneForty. The press allowed the site to go viral. This popularity resulted in the collection of 43,738 tweet ratings from 1,443 users in late December 2010 and January 2011.

Two kinds of findings were revealed by the study: the types of tweet categories that were liked or disliked, and the reasons why. André mentioned that the skewed sample population of the study — information-centric people or CNN readers — might influence the division of tweets into categories. Future research might try to generalize to other populations and Twitter as a whole.

The categories of tweets that were deemed worth reading were self-promotion (linking to something that the tweeter created), information sharing, and questions for followers. Presence maintenance (saying “Hello people!”), conversational, and “me now” (the tweeter’s current status) were the most strongly disliked.

Foursquare location check-ins were especially hated by users because of their irrelevancy. Every so often, someone “checked in” to a different city or even country, and users were typically annoyed by this. On the whole, the sample of Twitter users liked 36 percent of tweets, disliked 25 percent, and were neutral to 39 percent.

A tweet “being boring” was the most common reason to dislike it, accounting for 82 percent of why a tweet was rated NotWR. Other reasons included repeating old news (something as old as four hours was considered old news on Twitter), being cryptic (posting a link without a contextual explanation), or using an excessive number of # and @ signs.

When following celebrities or organizations, people looked for professional insights and were disappointed when receiving personal ones. News organizations were expected to only give a taste of the story in the tweet and not reveal all the information. Three components that increased the likelihood of a post’s popularity were information, humor, and conciseness.

Happy feelings were valued, as opposed to “whining,” which was disliked. The researchers advise tweeters to use a unique hashtag in questions so followers can keep track of the conversation.

The online persona that a person creates using social media may become more and more important for potential employers and personal connections. Last month, for example, a Twitter joke about “destroying America” (meaning having a good time) resulted in two British citizens being deported from the U.S., according to the Washington Post.