“My first rape and death threats came in 2005,” she says. Farrell wrote a blog post criticizing the US response to Hurricane Katrina as racist and was subsequently inundated with abuse. Since then, she says, the situation has worsened: “A decade or so ago, you had to say something that attracted opprobrium. That’s not the case now. Now it’s just every day.” She is extremely careful about which services she uses, and takes great care never to share her location online.
Death threats and online abuse aren’t the only online issues that disproportionately affect women, though. There are also less tangible harms, like algorithmic discrimination. For example, try Googling the terms “school boy” and “school girl.” The image results for boys are mostly innocuous, whereas the results for girls are dominated by sexualized imagery. Google ranks these results on the basis of factors such as what web page an image appears on, its alt text or caption, and what it contains, according to image recognition algorithms. Bias creeps in via two routes: the image recognition algorithms themselves are trained on sexist images and captions from the internet, and web pages and captions talking about women are skewed by the pervasive sexism that’s built up over decades online. In essence, the internet is a self-reinforcing misogyny machine.
For years, Facebook has trained its machine-learning systems to spot and scrub out any images that smack of sex or nudity, but these algorithms have been repeatedly reported to be overzealous, censoring photos of plus-size women, or women breastfeeding their babies. The fact that the company did this while simultaneously allowing hate speech to run rampant on its platform is not lost on activists. “This is what happens when you let Silicon Valley bros set the rules,” says Carolina Are, an algorithmic bias researcher at City, University of London.
How we got here
Every woman I spoke to for this story said she had experienced greater volumes of harassment in recent years. One likely culprit is the design of social media platforms, and specifically their algorithmic underpinnings.
In the early days of the web, tech companies made a choice that their services would be mostly supported by advertising. We simply weren’t given the option to subscribe to Google, Facebook, or Twitter. Instead, the currency these companies crave is eyeballs, clicks, and comments, all of which generate data they can package and use to market their users to the real customers: advertisers.
“Platforms try to maximize engagement—enragement, really—through algorithms that drive more clicks,” says Farrell. Virtually every mainstream tech platform prizes engagement above all else. That privileges incendiary content. Charlotte Webb, who cofounded the activist collective Feminist Internet in 2017, puts it bluntly: “Hate makes money.” Facebook made a profit of $29 billion in 2020.