Friday, February 3, 2012

How Activities of Individual Node Affect Its Neighbors

So far, we've mostly discussed the structure of networks and their "universal" properties. I was curious about how the actions of individual node affect its neighbors, and wanted to share a paper which studies how newcomers integrate to an existing network.

In "Feed Me: Motivating Newcomer Contribution in Social Network Sites"[1], the researchers from CMU and Facebook pose 4 hypothesis:

1. Social learning: Newcomers whose friends share more content will go on to contribute more content themselves.
2. Singling out: Newcomers who are singled out in content will contribute more content
3. Feedback: Newcomers receiving more feedback on their initial content will go on to contribute more content.
4. Distribution: Newcomers whose initial content is distributed widely will go on to contribute more content.

The study focuses on how early experiences impact newcomers' long term photo uploading behavior. Specifically, it builds models that predict the photo uploading behavior of 250,000 newcomers between their third and fifteenth week from their experiences in the first two weeks on Facebook.

The authors find support for hypothesis 1, 3, and 4. For social learning, the newcomers saw an average of 36.4 stories about their friends' photo uploads in their own News Feed. Every doubling of these photo upload stories was associated with a 6.1-10.7% increase in sharing.

On feedback, only 38% of the newcomers who uploaded at least one photo received a comment, and receiving even a single comment was associated with a 6.2% increase in subsequent photo sharing.

Results for the distribution hypothesis were modest, and when number of stories about the newcomer’s photos appearing in friends’ News Feeds was doubled, the expected number of newcomer photo uploads increased by 2.6%.

At first, I was mildly surprised at the relatively low increase in contribution, especially for hypothesis 1: social learning. Even when a newcomer experiences twice as many photo uploads from their friends, the sharing increases only by a single digit percentage. One possible explanation is that Facebook is already well known , and the amount of "social learning" required to understand what is technically and socially acceptable is minimal. These results suggest possible courses of action for designers of social networking sites, and it would be interesting to compare these factors for smaller, younger social networking services.


No comments:

Post a Comment