Anjana Susarla, from the School of Business at UW, is beginning to explore how social networks in Youtube (friends and subscribers) impact the rate in growth of “views” of freshly uploaded Youtube video.
The business motivation for studying this is that Youtube (and other user-influenced services or media) plays a heavy role in business today. Clear examples include network television and the music industry.
To study this effect, Susaria chose to explore the population of all Youtube videos that were uploaded under the “Music” category.
Variables
Each uploader was then mapped into two different social network graphs. One connected the uploaders by “friendship”, a formal Youtube tie. Another connected the uploader by “subscription”, producing a directed graph. With theory on social networks (sorry, I didn’t catch any specific references), Susaria could now assign a “centrality” value to each uploader based on their friendship ties or subscription in/out-degree. Susaria also tracked number of views, ratings, links, etc… all over time. All at a total of around two years of data collection.
Quirks to be accounted for
A particular challenge that Susaria had to face was that, unlike tradition diffusion studies, she had no way of connecting a “view” with a person. She didn’t know who viewed what video. Instead, she had to make diffusion and adoption insights from aggregate data: the total number of views that a video had to date.
Susaria also had to wean out
- effects that made a video popular due to influence outside of Youtube itself,
- video popularity effects that might be due to user heterogenity in tasks that lead to self-selection in group formation
Among her findings, she finds a tradition diffusion curve within the first 100 days of a video’s release that are based on the “centrality” of the uploader in his social network. Beyond those 100 days, external effects then come into play.
Susaria will continue her future work in this area, exploring external effects, sub-groups within the larger network, and content-related topics, such as video mash-ups.
Tags: Youtube
April 26, 2008 at 12:59 pm |
I found this research very interesting, as I am researching social networks for a graduate class. I would be interested in learning more about your research. How would a diffusion curve be applied to studying the popularity of posts within a social networking site, such as Ning.com?
April 26, 2008 at 7:15 pm |
Thomas, thanks for showing interest. I noticed that I mistyped Anjana’s name (which would make contacting her much harder…). I’ve corrected that and I suggest that you contact her about your questions. Here is her faculty site: http://bschool.washington.edu/faculty/faculty_detail.asp?ID=190