When the FCC asked the public to comment on its idea for internet “fast lanes, it received more than a million responses ranging from expert opinion and rants to form letters and, in one case, the entire text of War and Peace. While this reflected the high public interest in the topic, it also posed a problem as to how the agency — or anyone else — is supposed to slog through all that.
One way is to apply big data tools to the job and the Sunlight Foundation, a non-profit group that promotes government transparency, has just released its own findings based on machine learning and sentiment analysis. What did it find?
The most significant finding may be the lop-sided nature of the comments: fewer than 1 percent are “clearly opposed” to net neutrality, while around 5 percent of the comments represent opposition to regulation in general. Meanwhile, about two-thirds are actively opposed to paid fast-lanes or dividing the internet into fast and slow tiers. Other highlights:
- About 1,500 people employed the word “f*ckery* or “dingo” or “John Oliver” — a reference to a comedic sketch aimed at FCC Chairman Tom Wheeler
- 600 comments, or 0.08 percent, can cab be considered “expert submissions” by the Sunlight Foundation’s “back-of-the-envelope estimate”
- Common keywords included “pay to play,” “Netflix,” “Comcast” and “monopoly”
- Both Les Misérables and War and Peace were submitted as comments
- The FCC parsed just over 800,000 individual comments from the more than 1.1 million submissions
The most interesting insight, however, may be what the Sunlight Foundation’s tells us about how the comment sausage is made. Specifically, the group reports that 484,692 comments, or about 60 percent, are the result of organized “get out the comment” letter-writing campaigns organized by advocacy groups. As it turns out, that percentage is lower than usual the rate for similar situations that involve regulator asking for public comment (the number was reportedly closer to 75 percent for two recent processes involving the Keystone Pipeline and the IRS).
The report also offers a graph that suggests which interest groups are most effective at soliciting comments. CREDO Action’s activities reportedly led to 23 percent of all the “get out the comment” related submissions, while those of the Electronic Frontier Foundation produced about 18 percent of the submissions. Meanwhile, the activities of most active group on the other side, “Stop Net Regulation,” resulted in the fewer than 1 percent of submissions. Here’s the navigable graph, which also lets you read individual comments:
While the Sunlight Foundation’s efforts demonstrate the potential of tools like machine learning and semantic analysis, they are not always helpful. Like a recent NPR report on the FCC comments, the Sunlight Foundation’s findings provides an overview of the politics of the net neutrality debate — but don’t really surface the best or most convincing arguments to help the FCC or public decide on an answer.
Likewise, the data as presented doesn’t provide an obvious way to predict whether the comments will convince the FCC to rule against fast lanes — or if the million plus comments will just be another procedural box for the agency to check as it steams towards a decision. As such, it’s still hard to know how much these type of studies matter to the outcome of a policy debate, or if their main significance lies instead in demonstrating new forms of data analytics.