Journalists and Twitter: A Multidimensional Quantitative Description of Usage Patterns
Summary
This article details a quantitative analysis of the twitter usage of journalists, news organizations, and news consumers by Dr. Bagdouri. The study compares twitter usage metrics between these groups as well as between nationalities. To collect the data, the researchers gathered twitter profiles from http://media.info and used the Twitter API to collect over 13 million tweets associated with them. Using these tweets, the researchers measured 18 features aimed at assessing the perception, engagement, and delivery of each twitter account’s activity with its audience. These features were then analyzed using Kolmogorov-Smirnov and Welch’s t-test. From this analysis, the researchers concluded that not only did the three groups have very different communication styles, but there were significant differences across sections within each group.
Reflection
This article left me with several questions about the validity of the researchers’ methodologies. To begin, the researchers used a third-party website http://media.info to collect twitter account names associated with news affiliates, but did not include any evidence to support the credibility of this site. I did some digging around on the site, but I didn’t find any evidence to the contrary in what little time I spent. However, I still believe that this method is questionable. If the researchers wanted to observe journalists and news organizations, why did they not simply pull their data from one of the many circulation lists available online?
Going further, the researchers use a completely different method to collect the twitter accounts for Arab journalists. I’m not a dedicated researcher myself, but wouldn’t this difference in methodology introduce a bias in the data?
Finally, the researchers use an incredibly large collection of accounts and tweets (2.2 billion tweets) to analyze audience engagement. This collection process was inherently biased, however, and the researchers admit that the limitations could have been addressed through randomly sampling news consumers. From this I wonder: why did the researchers not randomly sample the journalist and news organization accounts in the beginning? Then they could have chosen one, consolidated source and reigned in the massive data project they chose.
