My Ph.D. advisor Dr. Joe Bob Hester came to me with a question: how do academic articles on the broad concept of social media cite each other? After we got talking about it, in addition to a traditional citation analysis, we thought we’d try a network analysis. The main point of discovery here was to try to figure out how articles cluster together and how connected the networks might be. We had some hypotheses: First we thought,  maybe articles that cite each other come from the same journals. Alternatively, we thought we might actually see journal articles citing each other based on discipline (i.e. advertising, public relations, business, psychology, etc.). Or perhaps citing could be found around a theory or a specific topic (i.e. word of mouth or uses and gratifications theory).

A novel way to see how articles go together is to see how they naturally cluster using a clustering algorithm. In the social network analysis tool we used, Gephi, we had several options.

For the first iteration, I let the computer cluster using the OpenOrd Layout. OpenOrd was designed to naturally discover clusters. You’ll see that the layout is busy, a lot of ties going all over. There does seem to be some “academic siloing” by discipline, but not a ton. This picture is messy and ultimately ambiguous.

Natural Clustering – OpenOrd – 3+ Mentions

The second time I adapted the computer’s clustering logic. This time we accounted for the attribute of discipline. ForceAtlas2 is the core developers of Gephi’s heralded own creation that emphasizes complimentaries or attributes that are the same. Because our designated attribute was discipline, it tried to cluster that way. ForceAtlas also pushes out nodes that seem to be outsiders to the network. It removed six nodes from the first iteration.

Qualitative Clustering by Discipline – ForceAtlas2 – 3+ Mentions

Here we see a much cleaner picture. In addition, we start to see some real academic siloing. Now granted, the computer’s algorithm naturally favors those clustering based on those that come from the same discipline, but we noticed that the network is much less confusing. There are fewer ties going across the entire network. Instead, ties are relatively local.

To explore our other two hypotheses, we recoded the attributes. Next, we examined journals. Was it possible that certain journals would cite other journals? We had a hunch that journals of similar disciplines would cite each other.

Qualitative Clustering by Journal – ForceAtlas2 – 3+ Mentions

Here we see that advertising and marketing seem to cite each other, while other social sciences and psychology seem to cite each other. While journals from each discipline send ties to other disciplines, the strongest ties appear to be between marketing and advertising.

Finally, we took a look to see how articles clustered together by topic. The premise here was that similar studies or theories would cite each other more than citing other topic areas.

Qualitative Clustering by Topic – ForceAtlas2 – 3+ Mentions

The clear topic here at the center of most studies looks to be word of mouth. That makes sense, as word of mouth is a novel theory for discussing the spread of messages. Uses and gratifications takes a distant second, followed closely by more general studies describing online social networks. Don’t be fooled by our clustering method here. This network is pretty well connected. That is to say that people are pulling different topical areas of social media when writing academically. The exception to this rule appears to be uses and gratifications, which seems to largely sit out at the end of the network. We also see stats as an outlier, but that’s likely due to all of the qualitative papers we are seeing.

Overall, our networks show a fair bit of cohesion. It looks like journal articles are citing across disciplines, topics and journals. We’re not claiming ubiquity here, there is some siloing by discipline, but overall, we see a well connected network. What we don’t see is diversity. Meaning, there aren’t a ton of topics/theories addressed here. In addition, disciplines that you might really expect to see in this mix, mainly public relations (and even advertising) are missing. In fact, we didn’t find one “big player” in public relations in any of these graphs. Instead, it looks like other disciplines such as marketing and psychology seem to run the table.

As always, you can reach me @chrisjvargo or @joebobhester if you have any questions.