spambot fake comments ethics

How NLG is reshaping ethics and responsibility

Late last week, Hackernoon’s Jeff Kao posted ‘More than a Million Pro-Repeal Net Neutrality Comments were Likely Faked’, an article and analysis that alleged that up to several million comments submitted to the US’s Federal Communications Commission (FCC) were generated by bots.

In his research, Kao focused on one campaign. That campaign, he estimated, had used mail-merge to disguise 1.3m comments as ‘unique grassroots submissions’. An overwhelming proportion of those comments were in favour of abandoning net neutrality.

On the other hand,” Kao continued, “comments in favour of net neutrality were more likely to deviate from a form letter and were much more numerous in the long tail. If the type, means of submission, and ‘spamminess’ of comments from both sides were equal, we would expect a roughly-even distribution […]. That is evidently not the case.”

Kao’s breakdown found that fewer than 800,000 of the more-than 22m comments could be considered truly unique. Of those unique comments, more than 99 per cent were in favour of keeping net neutrality.


It is important at this point to take a step back at this point and talk briefly about what net neutrality is. Net neutrality is a simple concept, yet one that sounds incredibly complex due to the unfamiliarity of the phrase. Good primers on what it is can be found here, here, and here. What it means essentially is that all data on the web must be treated equally by telecoms companies. So the website for an independent bookshop plays on the same level playing field as Amazon.

In 2015, the Obama Administration enshrined into law regulation that guarranteed net neutrality. That legislation ensured that internet access was classified as Title II, meaning that it could be regulated by a public utility—in this case, the FCC. In shorthand, it meant that providers were subject to open internet rules.

Since then, the world has changed and the 45th US president seems intent on undoing everything the 44th put into place. The current chairman of the FCC is Ajit Pai, who in a previous role was Associate General Counsel at Verizon, where he handled competition matters, regulatory issues, and counseling of business units on broadband initiatives. He outlined in April that the FCC intended to fully repeal Obama’s 2015 legislation, reclassifying the internet as Title I. This would mean that the Federal Trade Commission would take over regulation in this area and the present rules would be replaced by ‘principles’. These principles are much harder to regulate, reportedly, than rules. The only rule left at that point in this game would be that of transparency—the telecoms companies would have to inform consumers that they were choosing to throttle speeds.


Pai’s current justification for this is that net neutrality has been harmful for consumers, hamstringing investment by telecoms companies in infrastructure. However, this has been contradicted by the companies themselves, as seen in this video from Vice, which also offers a neat summation of the current state of play. The official vote regarding net neutrality and the new legislation is set to take place on 14 December.

There will be very-definite effects on the abandonment of net neutrality. Financially, getting rid of it means that larger organisations with financial clout will be able to dominate marketplaces, pushing out their smaller, independent competitors. It also means that telecoms companies will be able to push and divide their customers, creating a two-tier system for the internet. There are already moves in that direction post-net neutrality: as reported in ars technica, Comcast is already hinting at paid fast lanes following its repeal.


There are other implications. Any organisation that upsets a telecoms community may find their internet access cut short. We have seen this in a different context recently when Peter Thiel bankrolled a lawsuit against Gawker that ultimately saw the website forced to shutter.

The internet is currently open to everybody; abandoning net neutrality moves that influence to the grasp of a few, large corporations.

This type of large-scale attack, as seen on the FCC, is not uncommon in these times. And the effects have been profound and devastating. In 2016, hackers working for the Russian government, from a base in St. Petersburg, worked to try and sway the US Presidential election. Similar suspicions and smoking guns have tied that cyber-campaign in to interference with the UK’s Brexit referendum, an event that caused seismic shocks across Europe and has destabilised much of western politics. Even now, over a year from the Article 50 deadline, the UK remains and bitter and divided country. Further attacks, thought also to originate from Russia, came during the German federal election in September, which ended up bringing far-right group AfD into the German parliament. Even now, months later, Merkel is struggling to form a government and the possibility of another election is looming. There are fears, too, that Russian hackers worked their black magic around this year’s election in France and the independence referendum in Catalonia.

It is not yet clear who or what was responsible for the bots that attacked the FCC’s website. But it is only natural to look as to who would benefit from a repeal of net neutrality and to who has the resources to implement such a scheme.

Attacks like these are no longer isolated, nor are they anything other than commonplace. Despite Natural Language Generation (NLG) being open to all, there is still some organisation that is required. It cannot be accomplished by someone weighing 400lbs sitting on their bed.

Whichever organisation is behind these bots, the implication is clear: we are in a state now of perpetual cyberwar. Every shot fired across the bows in the last year has been a starting pistol for the next weapons race.


But it also shows the power of Natural Language Generation. This morning, we ran a test for one of our projects, checking some of the language code. The current speed of that project, according to my machine, is 24.19 texts per second. The texts also read pretty well.


The upshot? NLG is here to stay.

And if the last year has proven anything, it is that NLG is something that can be misused, especially if implemented in the shadows, as seems to have been the case. But NLG is not inherently a bad thing; it is, instead, simply a tool. The best comparison would be with a knife: you can use one to eat dinner or perform heart surgery. But you can also use it to rob or attack people.

At this point, it would be useful for a set of ethics to be developed by NLG companies, and we are glad to be leading the way. In our next blog, we are going to outline some of the issues and thoughts about what those ethics would be and where they could take us.