Having defined the basic pattern, one can start addressing the question: How does the Viewer get exposed to the Owner's photo?
1) Perhaps the Viewer might have stumbled upon the photo randomly,
perhaps via a group, or saw it on Explore, or any of many alternative
mechanisms not involving the social network.
2) Perhaps the Viewer had already added the Owner as a contact, and saw the photo in Owner's photostream.
3) Perhaps the Viewer browsed the contact's gallery of favorites.
The second and third options represent the central topic of this analysis: How does the social network affect exposure to content? In other words, to what extent does information percolate through the social network?
An important consideration is that perhaps the Viewer has not just one contact who faved the photo, but several contacts who faved it. Intuitively, a larger number of such contacts might increase exposure to the content.
Measuring exposure to content is not trivial. There is no way to assess, via the API (or by any other method that I know of) whether someone saw a photo, unless they leave a comment, a tag, a note, or fave the image. For the sake of simplicity, in this analysis I use faving as a proxy for exposure level.
To summarize the method, then:
1) I define a basic heterogeneous pattern.
2) I ask whether Viewer contacted Owner. [yes/no]
3) I ask how many people Viewer has contacted, each of which has faved the photo. [N]
4) I ask whether Viewer ended up faving the photo too. [yes/no]
Given a very large number of such patterns, each annotated with the two yes/no questions and the number of contacts, what can we learn about information percolation?
(See the next slide.,,)