Human rights NGO Amnesty International and international artificial intelligence company Element AI released a report today on all the Twitter harassment women endured in 2017, and their combined human-machine findings are pretty much what you would expect. Women get harassed a lot, with black women at an alarmingly higher rate. (Incidentally, all things women have been saying for years. )
The “shocking” report found that an abusive or problematic tweet was sent to a woman every 30 seconds and 84% of the time, that abusive tweet mentioned a black woman. Women of color were “disproportionately targeted,” wrote the study’s authors and it didn’t matter where they fell on the political spectrum, even Breitbart journalists were targeted.
“[We] have the data to back up what women have long been telling us – that Twitter is a place where racism, misogyny and homophobia are allowed to flourish basically unchecked,” said Milena Marin, the Senior Advisor for Tactical Research at Amnesty International, in a release on the findings.
The purpose of the study was not to police Twitter, but to ask Twitter to be more transparent about the algorithms it uses to detect abuse. Amnesty International has repeatedly asked Twitter to publish data on the abuse perpetrated on their platform, and the company’s failure to do so “hides the extent of the problem and makes it difficult to design effective solutions” noted the study’s authors. Let’s also be real for a moment and mention about another reason for this study -- to model just how well the combined efforts of man and machine work when it comes to content moderation. The joint study by Amnesty International and Element AI was a multi-pronged effort that began with Amnesty International’s Troll Patrol, a “unique crowdsourcing project designed to process large-scale data about online abuse.”
The Troll Patrol comprised of 6,500 volunteers from 150 countries who spent 2,500 hours analyzing 228,000 tweets sent to 778 notable women (like politicians and journalists) living in the UK and the US. The dataset volunteers built from these tweets was then extrapolated on by Element AI in order to comb through more than 15 million tweets for other abusive or problematic messages. “By crowdsourcing research, we were able to build up vital evidence in a fraction of the time it would take one Amnesty researcher, without losing the human judgment which is so essential when looking at context around tweets,” said Marin.
The computational data from this study backing up what women have been saying on the platform for years is great to have, as is an example of how well humans and machines can work together to identify hateful tweets, don’t get me wrong, but the study itself begs the question... why was it Amnesty International and an independent artificial intelligence research company that did this public study, and not Twitter? Twitter doesn’t even have as many dedicated moderators and employees as were volunteers in this study -- in 2017 the company was reported to have less than 3,600 employees. How advanced is their abuse-detecting robot anyway?
Twitter repeatedly dropping the ball when it comes to online harassment is not new news but what would be is if Twitter addressed how the platform’s own culture, and how people become popular on the service, can facilitate harassment. Easy engagement on Twitter usually means attacking, arguing or mocking random people for the entertainment of your followers and the platform generally rewards you for being shocking, divisive, controversial and angry more than it does when you are thoughtful or kind. In addition to Twitter as a natural breeding ground for harassment, there is a whole online content industry known as the “anti-SJWs,” not confined to Twitter, dedicated to hating on women and their progressive allies. Videos compilations of feminists to laugh at and ridicule regularly get millions of views on YouTube and anti-SJW internet personalities frequently harass women online for Internet points. Combine those factors with the "surge" in rightwing extremism under President Trump, anyone surprised black women face the brunt of online harassment hasn’t been paying attention.
