Last week SocialFlow took part in the 5th International Conference on Weblogs and Social Media. Even though ICWSM is a relatively new conference, it has become one of the premier venues for social scientists and technologists to gather and discuss cutting-edge research in social media. The conference hosted a good mix of academics along with leading industry data scientists, with a core of the presentations focusing on Twitter and Facebook data analysis.
I participated in two sessions. The first, a panel named Is Social Media Making News Generation and Consumption Better? moderated by Sihem Amer-Yahia, included journalists, editors, academia and industry representatives. The session covered broad themes such as the shift from broadcast to networked information sharing, information flows and the rising tension between editorialized and social recommendations. Mor Naaman of Rutgers made a case for the need to build innovative tools for journalists, while Marcus Mabry of the New York Times referenced the filter bubble, asking how we can make sure that the right voices are being heard, especially as media outlets cite an ever increasing number of blogs.
During the panel, I presented our take on information flows and audience analysis, including data from an upcoming study on Twitter audiences of prominent media outlets (to be published here next week). I focused on how we analyze networked audiences on Twitter based on groups of followers and mentioned a number of metrics we use to measure engagement levels. Should media outlets be focused on driving traffic (lots of clicks) or gaining brand trust (lots of Retweets)? Finally I described our study on viral information flows, noting that it has become virtually impossible for media entities to “break the news”! (see our previous ‘Breaking Bin-Laden‘ analysis)
My second presentation was a part of the Future of Social Web workshop, focusing on social data visualization. There I showed a number of visualization methods, showing examples from the web and questioning their effectiveness. We all agreed that visualization tools are helping researchers make sense of big data, but had different opinions on whether these tools are effective in providing regular readers with a better understanding of the news. One of my conference highlights was Sinan Aral‘s keynote on causality and content. Sinan outlined a thoughtful and incredibly thorough framework for thinking about social influence, highlighting the importance of distinguishing between homophily and influence. Effectively, when looking at a pair of friends, is one actually “influencing” the other and driving behavioral change, or do they behave similarly because of homophily (similar people are more likely to be friends). All sessions were video taped, and will be placed online in the near future. I highly recommend viewing Sinan’s talk.
Here’s a list of studies I found particularly interesting:
- A Machine Learning Approach to Twitter User Classification
- The Party is Over Here: Structure and Content in the 2010 Election
- Dimensions of Self-Expression in Facebook Status Updates
- Center of Attention: How Facebook Users Allocate Attention Across Friends
- Media Landscape in Twitter: A World of New Conventions and Political Diversity
- Why do People Retweet? Anti-Homophily Wins the Day!
- RT to Win! Predicting Message Propagation in Twitter
Full conference preceding can be found here. And finally, slides from my panel presentation are attached below: