As Twitter, Facebook, and other popular sites across the web continue to figure out how to fight trolling in the Age Of Trump, Google has taken its big step forward with the release of its Perspective API.
Created by Jigsaw (a Google-bases incubator) and Googles Counter Abuse Technology, Perspective is for publishers that are constantly dealing with the onslaught of horrible trolls. The tool monitors conversations happening in a forum in real time and assigns the comments a score relating to how “toxic” it considers the comment.
The program’s definition of “toxic” was created by asking internet users to rate comments on a scale from “very toxic” to “very healthy.” For reference, the definition of “toxic” in this context was, “a rude, disrespectful, or unreasonable comment that is likely to make you leave a discussion.”
In a chat with the Washington Post, Jigsaw president Jared Cohen noted that the team putting the API together collected millions of comments from the New York Times and additional comments from Wikipedia and online harassment victims to help them develop that toxicity scale.
It continues to be a work in progress, he notes, and as the API is now released into the wilds of the internet, more feedback would help the team improve it.
Jigsaw product manager CJ Adams said the purpose of Perspective was to empower publishers to “host robust debates and … help people stay online without having to read every bit of abuse hurled their way.”
There’s a great way for your average internet user to get a glimpse of the program, too. On the API’s website, there are some great interactive tools that show you how it works.
One tool displays dozens on comments on three topics: Climate change, Brexit, and the 2016 U.S. election. Using a slider tool, you can see how certain comments were filtered by toxicity. Slide all the way to the left and you get only the least “toxic” comments. Slide to the right, and you get everything, including the most “toxic” comments.
There’s also a tool where you can type your own trolling comment to see how “toxic” the API scores it. I tested it out using three phrases: one of my own, one of Donald Trump’s tweets, and a tweet sent to me by a troll on Twitter.
Though it’s in its early stages, sites can now request access to the Perspective API. Some of the publications already using it are the New York Times, The Guardian, and The Economist.
Ultimately, what the publisher decides to do with the API is entirely up to them, according to Jigsaw and Google. They can use the API to flag comments for moderators or even set up their forum in a way that allows each user to select the toxicity threshold. Or, they can even use the other tool and allow users to see how “toxic” their comments are as they type them.
While there will be continuing tweaks to the programming, it seems to be a solid step in the right direction, going beyond a “banned word” list that some forums use and allowing for nuance that a human moderator brings to the table.
And Jigsaw is definitely taking an optimistic approach to the API’s implementation, focusing on how it could evolve discourse rather than simply shut down harassment.
“Advances in machine learning suggest that machines might be able to facilitate the way we communicate,” says Lucas Dixon, Chief Research Scientist at Jigsaw in a press release, “and not only in terms of efficiency, but also in terms of the quality of our conversations too. Perspective has the potential to help people hear, understand, and resolve their differences.”