How a bot made 1 million comments against net neutrality look genuine

Hasn’t cleaned out his spam folder.
Hasn’t cleaned out his spam folder.
Image: Reuters/Aaron P. Bernstein
By
We may earn a commission from links on this page.

“Gathering and analyzing comments from the public is an important part of the Federal Communications Commission’s rulemaking process,” the American agency says on its website. But analyzing those comments increasingly means reading the thoughts of spambots.

Automated comments are now part of political reality: During last year’s US presidential race, a large proportion of tweets supporting both Donald Trump and Hillary Clinton came from automated accounts. These bots send messages en masse, originating from one source and usually conveying a particular ideology. Some are easy to spot. But others can be well-written and pass under the radar, making top-down campaigns look like they have more supporters on the ground. For this, comments on the FCC plan to repeal net neutrality rules provide an example.

Ajit Pai, the FCC chairman, announced last week that he intends to repeal Obama-era regulations that prevent internet providers from treating certain kinds of traffic differently, or charging more for it. On its website, the FCC solicited comments on his proposal.

Some pretty crude comment campaigns responded: One statement, that the Obama administration’s neutrality rules are “smothering innovation,” appears verbatim over 800,000 times in the FCC’s comment database. It’s not clear whether these comments came from a spambot or a highly motivated get-out-the-comment effort. What is obvious, though, is that all of these comments came from a single source.

Search for another pro-repeal comment, however, and there is exactly one result:

Dear Mr. Pai, In the matter of the future of the Internet. I strongly urge you to undo President Obama’s decision to take over broadband.

That uniqueness might suggest that it was written by a real person. But wait a minute, another “unique” result looks suspiciously like that one.

Dear Commissioners: Hi, I’d like to comment on network neutrality regulations. I want to urge the commissioners to overturn Obama’s order to control the Internet.

And another.

To the FCC: I’m concerned about an open Internet. I’d like to ask the commission to rescind The Obama/Wheeler plan to control the web. (Tom Wheeler was FCC chairman before Pai, under Obama.)

It keeps going.

Mr Pai: Hi, I’d like to comment on internet regulations. I strongly recommend the Federal Communications Commission to undo Tom Wheeler’s plan to control the Internet.

This is a more sophisticated campaign at work.

An analysis of all the comments in the database found over 1.3 million comments that followed this same basic structure. “It was like mad-libs,” wrote Jeff Kao, the data scientist who did the analysis, in a blog post explaining his findings. The fill-in-the-blank statement continued past those first sentences. Here is an example of a full comment:

Dear Mr. Pai, In the matter of the future of the Internet. I strongly urge you to undo President Obama’s decision to take over broadband. Citizens, as opposed to the FCC Enforcement Bureau, ought to use the services we desire. President Obama’s decision to take over broadband is a distortion of net neutrality. It reversed a light-touch approach that worked very, very well for two decades with broad bipartisan backing.

Every one of these 1.3 million comments has the same structure, with the bracketed terms changed in each case:

Dear [FCC]. I strongly [urge/recommend/ask] the FCC to [rescind/overturn/undo] the rules [set in place/laid down] by [Obama/Wheeler/both], which [take over broadband/control the internet]. [Normal people], as opposed to [elitist liberal bureaucrats], should be able to [use/purchase] the [services/applications/products] they want. The [Obama/Wheeler/both] plan is a [betrayal/exploitation/corruption] of [net neutrality/the open internet]. It [undid/reversed/broke] a [light-touch/market-based/pro-consumer] [approach/policy/system] that [worked/functioned/performed] successfully for [a long time] with [bipartisan support]

Each of the terms in brackets are replaced by synonyms, creating unique combinations. Those are just a few of the examples. In these comments, what is described as “a light-touch approach” has these variations and more:

  • “a market-based policy”
  • “a hands-off policy”
  • “a free-market policy”
  • “a market-based approach”
  • “a pro-consumer system”
  • “a light-touch policy”
  • “a pro-consumer framework”
  • “a hands-off framework”
  • “a free-market system”
  • “a free-market framework”

It may seem like a struggle to get to 1.3 million unique comments in this way, but the math is pretty straightforward. The statement has around 20 chunks of text that get expressed with different synonyms. If each of those has three possible options, then there are 3 to the 20th power combinations, or about 3.5 billion. It would be “incredibly simple from a code perspective” to write a program to compose and submit all of these comments, Kao told me via email. He said he could probably do it “in an hour or so.”

And while maintaining variation is easy for a computer, the chances that actual people submitted over a million variations of this highly specific theme, without overlapping once, are extremely low. Therefore, these comments are most likely the result of a spam campaign designed to make them look like the concerns of normal citizens.

The phrase “light-touch approach” adds some more intrigue to this story. When I was browsing these comments, it jumped out to me as sounding odd, so I figured it was just a lazy synonym. Then, I read Ajit Pai’s draft calling for repeal, and it appeared again: “For almost twenty years, the Internet thrived under the light-touch regulatory approach established by President Clinton and a Republican Congress.” In fact, it is a phrase Pai has been using since he took office. So these comments are adopting the specific language of the administration.

What does this say about the state of support for net neutrality? There were indeed similar campaigns on the other side of the spectrum—Kao identified a set of around 100,000 comments that advocated for keeping the Obama rules in place, and followed a template like the one above. But when he removed all of the clusters of similar comments, as well as all of the duplicates, 99% of the remaining comments supported keeping net neutrality.

But at least we now know what the bots think.