Well, no, Amnesty International isn’t sure. There are inherent flaws in their study that cause it to largely underestimate the abuse and harassment women receive on Twitter. Not the least of which is the fact that the study could not examine tweets that Twitter deleted or accounts that had been deleted. This means it couldn’t look at a sizable chunk of Twitter harassment. It could only look at the 91% of tweets that, despite reports, Twitter has not removed.
It’s a sizable study, validating the stories of harassment women tell. Trolls target women online at a greater rate than men. They also attack women of color, especially black women, at higher rates than white women. While this surprises no one who has been paying attention, the data backs up these stories and proves to networks how much work they still need to do to end censorship of female and non-binary voices on their networks.
Warning: This post will share some abusive Tweets. They’re examples of the harassment women face on Twitter.
In This Article:
The Data From the Study
“The violence and abuse many women experience on Twitter has a detrimental effect on their right to express themselves equally, freely and without fear. Instead of strengthening women’s voices, the violence and abuse many women experience on the platform leads women to self-censor what they post, limit their interactions, and even drives women off Twitter completely.”
– Amnesty International on how harassment hinders women’s right to free speech.
Impact on Women
Through extrapolation with AI provided by ElementAI, Amnesty International estimates a woman is harassed on Twitter every 30 seconds. Amnesty International collected 228,000 tweets targeting 778 women on Twitter. These women were politicians and journalists, representing a number of British MPs, every U.S. congresswomen, and a number of female journalists across the political spectrum. They found that 7.1% of tweets sent to women were harassment or problematic.
Racism Dwarfs Other Attacks
It’s clear that, while all women face more harassment on social networks than men, sexism is amplified by racism for these trolls, who are more likely to attack women of color.
Political Divide? Not by Much!
However, it’s interesting to see how the divide differs. Liberal journalists are more likely to face harassment than conservative journalists. Meanwhile, conservative politicians are more likely to see harassment than liberal women. As a result, they’ve leveled out, for the most part.
The Flaws in Amnesty International’s Study
“Troll Patrol means 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.”
-Milena Marin, Senior Advisor for Tactical Research at Amnesty International
Only Accounts for Tweets Twitter Hasn’t Removed
Honestly, these are almost tame compared to some of the examples of hateful speech and harassment I’ve seen.
Twitter is notoriously bad at catching hateful content, harassment, and abuse. In fact, of the 2.8 million reported tweets, Twitter only acted on 9% of them. As a result, 91% of tweets that users found objectionable are still on the platform. These were the only tweets that Amnesty International could analyze. As a result, they may have underestimated the volume of harassment reports women receive.
Missing Data on LGBTQ Harassment, Specifically of Trans Women
LGBTQ people face a lot of harassment online. They’re also underrepresented in government and journalism. Because of this. Amnesty International’s sample group largely excluded LGBTQ women. This is especially true of trans women, one of the largest targets of harassment on Twitter. However, these women were highly underrepresented in this study. Had trans women, lesbians, and bisexual woman been better represented in this study, the results would have been different. Amnesty International would have been able to report on the effects of a woman’s sexuality in relation to the harassment she faces online. Because they had so few people who are lesbian, bisexual, transgender, or queer, they were not able to report any findings. This cut out a large group of people who are frequently harassed, and likely caused their AI models to drastically underestimate harassment the average woman faces.
Amnesty International should have looked outside of politicians and journalists. They should have included social influencers, noted feminists, and actresses. Future studies will need to include LGBTQ people better. As you’ll see below, Amnesty International included Pink News, a popular LGBTQ network in the U.K., but it’s more popular with gay men.
Fake News is Not Conservative News
Conservative doesn’t always mean hateful, bigoted, or fake news. So why did Amnesty International use only those sources?
For journalists, Amnesty International looked at Daily Mail, Gal Dem, the Guardian, Pink News, the Sun (U.K.), and Breitbart. Obviously all attacks on women for the sake of their gender are bad. In an argument, people should attack the idea, not the person. But when people get angry, they often forget how to be a decent human being. This behavior is unacceptable, I’m not making an excuse for it. But fake news can inspire anger more than actual news. The Daily Mail is a conservative tabloid and Breitbart is an extreme right-wing propaganda network. Finally, there’s the Sun, a slightly more reputable right-wing network. However, not a single one of their conservative news sources are factually based or considered low on bias. They have questionable sources, and frequently report non-factual content, according to mediabiasfactcheck.com. Not a single conservative source has a factual, non-biased, or even right of center rating.
I believe that the anger that fake news and harmful, often racist, homophobic, and transphobic propaganda can generate more strong emotions than simple political disagreements. Hate speech is bound to receive more anger than normal discussion. Because the liberal sources were from legitimate sources, rather than extreme leftist sources, that same level of frustration is only on one side of this dataset. Therefore, Amnesty International introduced bias. They selected extreme right wing organizations that were bound to draw more ire than legitimate news sources. Perhaps this is why it seemed as though there were equal amounts of harassment on the right as there is on the left.
Racism in Elections, Professional Environments, and reduced racial diversity in the U.K.
Elected officials are predominantly Christian, straight, white, and male.
Racial diversity is significantly lower in the U.K. than it is in the U.S. This increases the degree of uncertainty in Amnesty International’s findings on race. In fact, women and people of color are extremely underrepresented in the U.S. government as well. Unless Amnesty International was sure to examine the same number of tweets from each racial group, their data isn’t as clear as it could be. We can extrapolate from this data, but this lowers our degree of certainty. In fact, look at the chart in the “Racism Dwarfs Other Attacks” section above. Note the size of the black bars for degree of uncertainty when it comes to all races buy white. Amnesty International’s data sources were disproportionately white because they went to a country that is less racially diverse than the U.S. and went to professions like politics that have extreme racial biases.
Women are excruciatingly underrepresented by the U.S. government.
Lack of Religious Diversity
Again, due to xenophobia and religious based biases, few popiticians are anything but Christian in the U.S. This is why Amnesty International was not able to report on whether or not Jewish, Muslim, Pagan, or atheist women are more likely to face abuse online. By limiting their dataset to specific women in these heavily biased industries, Amnesty International made their data as homogeneous as they could have.
Amnesty International Assumed Races
This is admittedly a bit of a nitpick, but the races of the women involved were based on a number of factors. Not at the top of the list? Asking the women about their family’s history and heritage. This could lead to problems. For example, say you’re only looking into people’s names. Look at embattled game designer, software engineer, and former political candidate Brianna Wu. Going solely off of her last name, you might assume she’s of east Asian descent. Her husband is, but she is not. By not allowing the subjects to label their own datasets, they introduced a potential source of error. Did Amnesty International make a mistake? Unlikely. They apparently used sourcing via Wikipedia articles, which certainly isn’t perfect.
This is the second way that Amnesty International failed to appropriately consider race during this study. It doesn’t invalidate their findings, but we can’t be sure of how much it influenced them.
Extrapolation and Models
Extrapolating from these incomplete, mostly straight, white, and cisgender datasets means they make assumptions about all female Twitter users based on a subset that is less diverse than Twitter’s actual users. Basically put, Amnesty International’s study was biased to make Twitter seem nicer than it actually is for the average female Twitter user. Amnesty International’s shocking results are likely a dramatic underestimate.
Future Studies and Twitter Involvement
Jack Dorsey, Twitter’s CEO
Future studies on harassment on Twitter should look at this study as an excellent stepping stone. It sets expectations, builds models that are more useful than Twitter’s own methods for detecting abuse, and, unfortunately, showed a number of areas for improvement. Here are a few ideas to keep in mind if you’re planning a similar study on Twitter harassment:
- Involve Twitter, if you can. They have data on deleted tweets you can’t get otherwise.
- Use a diverse dataset. Ensure you have representation from a variety of people on Twitter. Women face harassment for a variety of reasons, gender identity, sexuality, religion, race, and more. The sources of these tweets must be a cross section of Twitter’s actual users.
- Ensure the sources you use on the sides of the political spectrum are equivalent.
What Social Networks Need to Do
This study was far from perfect, but it revealed very real data about a problem that plagues women online. Allowing harassment and hate speech on social networks silences discussion. It removes variety, and forces those who are either targets with that harassment or uncomfortable witnessing it off the platform in question. Allowing hate speech is actually anti-free speech. People won’t participate in a network like that. Basically, social networks need to evolve or die.
Every 30 seconds, at least one woman faces harassment on Twitter. This article likely took you around 4-8 minutes to read. During that time, as many as 14 women where harassed. Likely more, due to the inherent underestimation in Amnesty International’s data.
Tell me, if you were on a bus, train, or in line at the grocery store, and every 30 seconds, you witnessed someone harassing a woman, would you do something about it?
Now’s your chance.
Demand change from Twitter. Ask them to work with organizations like Amnesty International to improve their platform. And, yes, be ready to boycott. With people like Jack Dorsey, Twitter’s unsympathetic CEO, empathy alone won’t make decisions. You have to hit his wallet.
Twitter needs better reporting systems. They need to ban those who abuse people on their platform, not allow 91% of reported tweets to go unchecked. They need to understand that harassment and threats online have a very real effect on people, and can even represent a real danger to their lives. Twitter needs to act, and we may have to force them to do so. We’ll take Twitter to an appropriate level of decency, even if we have to drag them kicking and screaming.
#ToxicTwitter #BoycottTwitter
Sources:
- Amnesty International [1], [2], [3]
- Philip Bump, Washington Post
-
Laure Delisle, Alfredo Kalaitzis, Krzysztof Majewski, Archy de Berker, Milena Marin, Julien Cornebise, Methodology Note for Amnesty International Study
- Rachel England, Engadget
- Melanie Ehrenkranz, Gizmodo
- U.K. Ethnic Groups (Wikipedia)
- U.S. Census