A guest post by Jeff Hemsley, who has co-authored with Karine Nahon a new book titled Going Viral.
In Going Viral (Polity Press, 2013) we explore the topic of virality, the process of sharing messages that results in a fast, broad spread of information. What does that have to do R, or the R-bloggers community? First and foremost, we use the R-bloggers community as an example of the role of interest networks (see description below) in driving viral events. But we also used R as our go-to tool for our research that went into the book. Even the cover art, pictured here, was created with R, using the iGraph package. Included below is an excerpt from chapter 4 that includes the section on interest networks and R-bloggers.
<excerpt from chapter 4 of Going Viral: Interest Networks>
In the last section we focused on what made content remarkable to people. One set of these factors was information characteristics (novelty, resonance, quality and humor). Another important information characteristic related to viral events is the interest around the topic of content that connects people together. Researchers have found that the interestingness of content, as rated by study participants, is certainly one factor in whether or not people will share links (Bakshy et al. 2011). Recall that in chapter 2 we said that viral events can facilitate the creation of interest networks, which we said were temporally bound, self-organized networks where membership is based on an interest in the information content or an interest in being included in the interest network of others. What this means is that a viral event may emerge out of people sharing content that is interesting to them even if the content doesn’t have high production value, isn’t funny, doesn’t resonate with them and doesn’t leave them with positive emotional feelings.
How do interest networks work? Many of us are passionately interested in topics that the people we are close too are ho hum about. Here is an example. Jeff is a passionate R programmer. According to the R website, “R is a free software environment for statistical computing and graphics”. He enjoys solving math and programming problems that allow him to make visualizations of data, particularly social networks. While most of the people he is close to think the visualizations are aesthetically appealing, none of them are interested in how they are made. But some people are interested.
We’ll use a video that Jeff posted on YouTube to explore interest networks a little further. In 2012 Jeff made an animation of a Twitter network. The links faded out over time which was a new way of visualizing dynamic social networks. Frankly, the animation won’t hold most people’s attention for the entire 2 minutes, and most people that ran across the link would not click on it. Unless they had a very specific set of overlapping, or related interests. Since R is used in so many different kinds of analysis – everything from biostatistics, data analytics, finance, general statistics and even analyzing speech patterns – we would expect a relatively small subset of them to be interested in network analysis, and an even smaller subset to want to watch an animation of a dynamic social network. While Jeff is a dedicated R programmer, he knows few people in the broader community who share his interests in network animation, so when his video went viral he was quite surprised. To be more exact, a “how to” blog post he wrote, with the video embedded as an example, went viral in the R community, which caused the video to go viral and get shared into groups that had an interest in the content.
Here’s how it happened. Jeff wrote the blog post with the video and detailed instructions so that other R users could create their own network animation. He submitted the post to R-bloggers, a well-known blog about R programming (see original R-bloggers post here). In the R community R-bloggers is a hub in that it connects R users who read and write about using R. The visitor traffic for R-bloggers is roughly 10,000 visitors a day, substantially higher than the traffic for SoMeLab.net, where Jeff posted the article. R-bloggers in this context acts as a network gatekeeper in that it connects users and networks who wouldn’t otherwise be connected. Jeff’s video went viral within the R community, an interest network.
Recall that in chapter 2 we said that the concept of virality was scalable, so a few hundred views could qualify as viral if the process by which it spread fit the criteria of reaching a critical mass within the interested community and being shared socially. If no one shared the video and all the views came from people who watched the embedded video on R-Bloggers, we would call this a kind of broadcast. But people did share it. Using YouTube analytics we found that 52% of the views came from the R-bloggers site and 13% came from SoMeLab.net, with Google, Facebook, Reddit, and Twitter making up 27% of views and the remaining views coming from YouTube sources. Using sites like Bitly.com and Topsy.com we were able find Tweets and other social media shares where users had posted or tweeted a link to the video into their own networks, and in some cases retweeted or reshared these links. For example, we found traces of the video in topic areas such as data visualization groups and forums about general statistics. Clearly beyond our expectation of a narrowly defined set of overlapping topics: network analysis and data animation. From our analysis about a quarter of the views the video received came from social sharing of some sort. We also found that when two key R users, who had affiliations with other interest communities, shared the video, it spread into these other groups as well.
This highlights that when people find content remarkable in a topical area that they are interested in, a viral event may emerge as the content hops from network to network. We call this an interest network because it is based on a topic of interest along an ephemeral network constituted around a specific topic. When we trace these networks we see how people are connected through information diffusion. Since the network is based on the movement on content, once the content stops flowing the network ceases to exist. However, the paths, or links, by which the content flowed, are often known or discoverable. This means people that are weakly connected, or not connected at all may end up having conversations about the content. After Jeff’s blog post and video went out he answered several emails and followed up on a few comments on the blog. Some of these have led to conversations.
Thus, viral content can facilitate conversations, which, over time, can form more durable connections. The reach of viral events – its ability to hop from one network to another – means that people may be randomly exposed to ideas, information and other content that otherwise might never have reached them. As a result they may find that they unexpectedly share interests with people they only casually know, or they may find that there are existing communities who share the same interests they have. We think that repeatedly invoked interest networks, over time, may bring people together who, by discovering they share topics of mutual salience, may form more durable, influential and actionable networks.
Just to summarize, here are the key characteristics of interest networks: 1) since messages are forwarded from person to person, interest networks come into existence along the links that connect people who are interested in the same content; 2) these networks are ephemeral, meaning once the message ceases being forwarded, that specific interest network may no longer exists; 3) if many, topically similar viral events invoke interest networks with the same members, more durable connections may form and, 4) these networks are bound by the collection of people who view the message, whether they re-share it or not.
You can contact me through Twitter at @JeffHemsley
More information about the book can be found here http://eKarine.org/goingviral
Bakshy, E., J. M Hofman, W. A Mason, and D. J Watts. 2011. “Everyone’s an Influencer: Quantifying Influence on Twitter.” In Proceedings of the Fourth ACM International Conference on Web Search and Data Mining, 65–74. ACM.