Google Looks to Curb Hate Speech with Jigsaw
January 6, 2017
No matter how advanced technology becomes, certain questions continue to vex us. For example, where is the line between silencing expression and prohibiting abuse? Wired examines Google’s efforts to walk that line in its article, “Google’s Digital Justice League: How Its Jigsaw Projects are Hunting Down Online Trolls.” Reporter Merjin Hos begins by sketching the growing problem of online harassment and the real-world turmoil it creates, arguing that rampant trolling serves as a sort of censorship — silencing many voices through fear. Jigsaw, a project from Google, aims to automatically filter out online hate speech and harassment. As Jared Cohen, Jigsaw founder and president, put it, “I want to use the best technology we have at our disposal to begin to take on trolling and other nefarious tactics that give hostile voices disproportionate weight, to do everything we can to level the playing field.”
The extensive article also delves into Cohen’s history, the genesis of Jigsaw, how the team is teaching its AI to identify harassment, and problems they have encountered thus far. It is an informative read for anyone interested in the topic.
Hos describes how the Jigsaw team has gone about instructing their algorithm:
The group partnered with The New York Times (NYT), which gave Jigsaw’s engineers 17 million comments from NYT stories, along with data about which of those comments were flagged as inappropriate by moderators.
Jigsaw also worked with the Wikimedia Foundation to parse 130,000 snippets of discussion around Wikipedia pages. It showed those text strings to panels of ten people recruited randomly from the CrowdFlower crowdsourcing service and asked whether they found each snippet to represent a ‘personal attack’ or ‘harassment’. Jigsaw then fed the massive corpus of online conversation and human evaluations into Google’s open source machine learning software, TensorFlow. …
By some measures Jigsaw has now trained Conversation AI to spot toxic language with impressive accuracy. Feed a string of text into its Wikipedia harassment-detection engine and it can, with what Google describes as more than 92 per cent certainty and a ten per cent false-positive rate, come up with a judgment that matches a human test panel as to whether that line represents an attack.
There is still much to be done, but soon Wikipedia and the New York Times will be implementing Jigsaw, at least on a limited basis. At first, the AI’s judgments will be checked by humans. This is important, partially because the software still returns some false positives—an inadvertent but highly problematic overstep. Though a perfect solution may be impossible, it is encouraging to know Jigsaw’s leader understands how tough it will be to balance protection with freedom of expression. “We don’t claim to have all the answers,” Cohen emphasizes.
Cynthia Murrell, January 6, 2017
Comments
-
- Subscribe to Beyond Search
Feature archive
News archive
-
