Battling for the Net: Big Data and Net Neutrality Activism

• In Event: ICA Interactive paper/Poster Session I
  In Interactive Paper Session: Computational Methods Interactive Poster Session

Abstract

In 2014, Federal Communications Commission (FCC)’s Chair Tom Wheeler proposed revisions to net neutrality, a regulatory concept that governs the internet, which led to nearly 4 million comments submitted to the organization by concerned citizens. Constituting the single most commented issue in the FFC’s history, the organization found itself unable to analyze the “big data” comprising the “Open Internet Docket,” the public file pertaining to net neutrality. As such, the FCC turned to data scientists at large on the internet, publicly soliciting help on their website blog and publishing the Open Internet Docket data corpus. In response to the FCC’s call, this work uses latent Dirichlet allocation (LDA) to create topic modeling of the entire corpus of comments in the Open Internet Docket (N ≈ 3.7 million). With LDA, 1000 topics were generated and manually coded by two researchers, revealing 47 salient topics containing the most unique and telling words pertaining to net neutrality. The Open Internet Docket comments contain geographic data by state, city, and zip code, so the 47 significant topics were used to run multiple linear regressions with US Census data from the American Community Survey (ACS). Results reveal that the majority of comments were concentrated in California, Texas, New York, Florida, and Illinois; linear regressions show a distinctive use of word choice in comments according to household income; and political affiliation and household income are the strongest predictors for rates of comments submitted to the FCC among the 100 most active zip codes.

Authors

• Rogelio Alejandro Lopez, U of Southern California

• Larry Zhiming Xu, Annenberg School for Communication and Journalism, U of Southern California