1. Dynamics of Information Virality Karine Nahon, Jeff Hemsley and Shawn Walker karineb@uw.edu http://eKarine.org http://twitter.com/karineb jhemsley@uw.edu stw3@uw.edu Presentation to Google – March 2011
2. retroV - Who Are We?http://retrov.org Karine Nahon Jeff Hemsley Shawn Walker MuzammilHussain
3. Project Scope Understand the dynamics of information distribution on networks with respect to new media channels, like blogs and videos. Understand the power relationships between the stakeholder groups and their influence in information distribution
4. What does it mean viral? “network-enhanced word of mouth” (Darper, 1977) “a communication and distribution concept that relies on customers to transmit digital products” (Helm, 2000) “a type of marketing that infects its customers with an advertising message, which passes from one customer to the next like a rampant flue virus” (Montgomery, 2001)
7. What makes Information Viral? Top-down Approach Virality is a process governed by reliance on powerful gatekeeping nodes or elite The tipping point (Gladwell, 2002) Opinion leaders (Katz and Lazarsfeld, 1955) Bottom-up Approach Gatekeepers play an important but not crucial role (Herring, 2005) Situational factors determine the virality. Tail nodes and hubs are the same. (Watts and Dodds, 2007)
8. Political Blogs - Literature Cass Sunstein’s camp Homophily (Adamic and Glance, 2004; Hargittai et al., 2008; Lawrence, 2010) Fragmentation (Sunstein, 2001, 2008) Polarization (Sunstein, 2001, 2008; Hindman, 2008) Power law (Farrell and Drezner, 2008; Karpf, 2008)
9. Political Blogs - Literature YochaiBenkler’s camp More Choices Participation Deliberation (Benkler and Shaw, 2010; Woodly, 2008)
10. Gaps in the Literature Focusing on top-blogs only (Hargittai et al., in our dataset 24% of videos were not linked by top-blogs, and top-blogs linked to only 13% of the viral videos) Static linking models at one point of time Linking: blogs to blogs Very rare to see mixed methods (qualitative and quantitative)
11. The Research Questions What are the relationships between different types of blogs and political viral information diffusion? What is the difference between elite blogs and tail blogs in that process? Are there other types of blogs worth our attention as scholars? What would a life cycle that represents virality in this context looks like?
12. Data, Data… and Data Five datasets Viral videos list (3 categories: political, election and general resulted in 120 videos) Viral video daily-view data Blogs linking to viral videos Traffic data for the blogs Identifying four types of blogs Blogs (n=9,765), Posts (13,173), Videos (n=120) Between March 2007-June 2009
13. Four types of blogs Elite Blogs Huffington Post & Daily Kos Top-political Blogs Elite political blogs Top-general Blogs Blogs with more than 250,000 daily unique views Tail Blogs Remaining blogs
15. Methodology - Multiple Regression VIEWS = ELITE + TOP-POLITICAL + TOP-GENERAL + TAIL + CONTOLS +ε VIEWS = ELITE_t + ELITE_t1 + TOP-POLITICAL_t + TOP-POLITICAL_t1 + TOP_GENERAL_t + TOP_GENERAL_t1+ TAIL_t + TAIL_t1+ CONTOLS +ε Since our primary goal is to present a life–cycle of blog-post timing in the political information diffusion process, each independent variable group contains two variables: A count of links from blogs in that category to a given video in a given day. For example, ELITE_t, would represent all the links from the elite blogs to a given viral video on a given day t. A one day forward-lagged version of the link count variable to the views. This variable associates links from day t+1 (tomorrow) to view counts of day t (today). For example, ELITE_t1, would represent all the links from the elite blogs on day t+1 to view counts for a given viral video on day t.
17. If [Variable]t is positive, it means that blogs of this type post links to a video on the day of the peak. If [Variable]t is negative, it means that blogs of this type post during the decline, the link count is increasing while daily views is decreasing. If [Variable]t1 is positive, it means that blogs of this type post on the day after the peak.
24. The Underlying Concern Determining the conditions that are sufficient for creating, or maintaining, stable democratic practices and examining what exists today
25. Research Questions: Do political blogs of the same political inclination tend to link to the same content? If they do, to what extent? Additionally, in cases where cross-ideological linking occur, what is the nature of that linking? Do political blogs follow a bandwagon effect, that is, is there a positive relationship between the likelihood of blogs linking to content and the popularity of that content? Are blogs with a higher number of comments more likely to refer to viral videos?
32. Types of Homophily and cross linking Type 1 – Videos level – when a video of a certain political inclination receives links from a blog of a similar (homophily) or dissimilar (cross-linking) inclination. Type 2 – Blogs level – when two blogs of a similar (homophily) or dissimilar (cross-linking) inclination link to the same video. Type 3 – Posts level – when a blog post of a similar (homophily) or dissimilar (cross-linking) inclination links to a video.
33. Homophily and Cross-Linking: Type 1 Generally: 76% Homophily, 23% Cross-linking Homophily: 75% liberals, 17% conservatives Cross-linking: 17% liberals, 46% conservatives (Type-1 - Videos level – when a video of a certain political inclination receives links from a blog of a similar (homophily) or dissimilar (cross-linking) inclination. )
38. Washington Monthly Newsbusters Patterico's Pontifications Gateway Pundit Powerline Blog Crooks And Liars Hot Air Ann Althouse Open Left Five Thirty Eight Jawa Report IMAO Talk Left Red State MyDD Firedoglake Pandagon Huffington Post Ezra Klein DailyKos Shakespeares Sister QandO Talking Points Memo Juan Cole America Blog Digby Feministing Sadly, No! Homophily and Cross-Linking: Type 2 Type 2 – Blogs level – when two blogs of a similar (homophily) or dissimilar (cross-linking) inclination link to the same video.
39. Homophily and Cross-Linking: Type 3 Type 3 – Posts level – when a blog post of a similar (homophily) or dissimilar (cross-linking) inclination links to a video. Cross-linking: 62 posts (21% of blog posts) linking to 15 videos
Over all Top videos, top 100 political videos, and top 100 election videosTop unique YouTube URLDaily viewing data for each video125 videos; 65 with complete viewing data
Over all Top videos, top 100 political videos, and top 100 election videosTop unique YouTube URLDaily viewing data for each video125 videos; 65 with complete viewing data