‘New York Times’ Uses AI From Google To Remove ‘Toxic’ Comments


The New York Times will soon enable the comments section for more of its online articles due to a successful partnership with Google. The publication is using machine learning technology developed by the search giant to identify and remove “toxic” comments, reducing the moderation efforts required.

Trolls who write deliberately provocative and offensive posts on articles have long been the bane of newspaper websites. The problem has become so large that many publishers have given up on comments altogether, avoiding the issue of hiring human moderators by deleting the comments section.

The New York Times previously included the comments section on around 10 percent of its daily posts. This proportion was manageable by its staff, enabling editorial teams to monitor for obscene or inflammatory content. The paper is now increasing the number of posts with comments to 80 percent, courtesy of Google’s artificial intelligence systems.

Image of a humanoid robot with its hand raised to its face in a "thinking" pose
[Image by Thinkstock/PhonlamaiPhoto]Featured image credit: PhonlamaiPhotoThinkstock

The technology assesses each new comment to determine whether it’s likely to be a considered response to the story. It is trained to rank comments based on the likelihood of a staff member making a certain judgment, such as allowing the comment to remain or immediately deleting it.

When it’s been processed by the AI, each comment is added as a dot on a histogram graph visible to human moderators. The histogram is sorted to include reasons why the comment could be removed, such as a lack of thoughtful information or a provocative tone. Currently, comments are scored for obscenity, toxicity, and their overall probability of being rejected by a human.

The Times moderators can then review batches of comments at a time, such as those with a less than 20 percent chance of being rejected. The staff member can skim over each item to verify the AI’s judgment. The entire batch is then approved with a single click.

Image of a robot hand typing on a laptop keyboard
[Image by Thinkstock/iLexx]Featured image credit: iLexxThinkstock

The process enables the Times to moderate a hugely increased volume of comments each day. After successfully testing the system, the company is now rolling it out more broadly. All of the paper’s top stories will have comments enabled going forward. The company said it wants to “reimagine what it means to ‘comment’ online,” noting it is one of the first to try such an experiment.

“The Times is developing a community where readers can discuss the news pseudonymously in an environment safe from harassment, abuse and even your crazy uncle,” said Times community editor Bassey Etim. “We hope you join us on the journey.”

The comments section is available for the majority of New York Times articles from today. It’s likely the AI-powered approach to commenting, provided by Google subsidiary Jigsaw, will see adoption by other media organizations in the future. By involving machines, comments could return to the web, enabling considered and constructive debate of topical issues.

[Featured Image by JIRAROJ PRADITCHAROENKUL/Thinkstock]