Gilad Lotan, 2012 Conference
ANN-SONIC Fourth Annual International Seminar on Network Theory:
Networked Social Movements and Network Theory
Analyzing Information Flows and Networked Audiences:
What We Learn From Data
Presentation summary by Sandra Evans and Alex Leavitt
Gilad Lotan is the VP of Research and Development at SocialFlow, a New York City company that uses science and real-time data to help businesses earn greater attention and engagement on Twitter and Facebook. Previously, Gilad served as a program manager at Microsoft’s FUSE labs. Past work includes ‘Retweet Revolution’, visualizing the flow of information during the 2009 #IranElection riots, and a 2011 IJOC study investigating the relationship between mainstream media and social media channels during the Tunisian and Egyptian revolutions. Gilad’s work has been presented at TED, IXDA, Summit Series, Berkeley BCNM, Boston Book Festival, and published at HICCS, CHI and ICWSM.
Gilad Lotan was unique among the conference speakers as a representative from industry, not academia. His work at SocialFlow provides an interesting application of social network analysis to a marketing setting. Much of his work focuses on mining Twitter data in real time and focuses on networked audiences. One example of this type of work involves analyzing Twitter patterns to assess when and how a magazine, like the Economist, should strategically tweet about stories. His talk consisted of the following topics: product, audiences, and information flows and networked actors.
The product he works on addresses the following:
- Enabling predictions about how data spread through networks
- Focusing on news content in particular and how news organizations can choose what to publish and when
- Modeling audiences
- Mapping flows of information
One example of this analysis was a visualization of diurnal patterns within tweets about key words such as “coffee” and “tired”. These patterns of word usage also pointed to that fact that US time zones tend to dominate certain patterns of Tweets. He also demonstrated how Twitter data can be used to identify key events like major sporting events, and how SocialFlow can compare trends such as organic decline vs. algorithmic decline (as seen in spam networks). Understanding Twitter algorithms is also a way to make sense of phenomena like that fact that OWS hashtags have not trended.
- Visualizations of Twitter data can show how Al Jazeera audiences vs. Fox news audiences communicate about events
- SocialFlow can track changes in clusters of terms over time and can track co-occurrences of terms
- SocialFlow can track what people click on and news traffic – even words from articles
- One example was a descriptive cluster diagram on Kony 2012. This diagram provided an overview of how key locations (school-based communities in parts of the US connected with Invisible Children) played central roles in disseminating the Kony 2012 video on Twitter
Information flows and networked actors:
- Other projects look at the relationships between media outlets and individuals. A recent academic research paper used both content analysis and network analysis to map out Tunisia and Egypt twitter networks. The goal was to understand roles of actors (such as bloggers and journalists) in spreading information.
- A key takeaway was that their content analysis method was effective. The network involved many individuals rather than organizations who have greater reach (in terms of Twitter followers for example). Twitter made the role of these individuals more visible.
Q: A key discovery in this Tunisia/Egypt paper is identifying networks of individuals – a defining characteristic of the current movement; individuals connecting to individuals as conveyors of information and information flows.
Gilad Lotan: Yes, personal networks and flows are interesting in how connections between people affect flow.
Q: Are there different ways of getting more granular in terms of tweets?
dana boyd (co-author on paper with Gilad Lotan): We need to scale next time perhaps by using Mechanical Turk. It was challenging coding where people were; many were multi-located and this is something to explore (placeless or global people).
Q: How did you define journalist vs. blogger?
dana boyd: Journalists were those identified with a clear news organization; identifying as a blogger was based on self-identification.
Q: What about sentiment analysis?
Gilad Lotan: Everyone’s trying to do this but it’s extremely hard. Co-occurrence is a way to get at something similar. Sarcasm is really hard. Another issue is multiple languages.