Safe Spaces Tweets

From Gender and Tech Resources

Questions missing

Bots Against Trolls

If you use Twitter, 4chan or any online forums where comments are allowed, you might have noticed that they tend to not be very safe spaces. Kathy Siera says there is a “koolaid point” for women and queers who start to gather a following and be listened online. At this point, a certain group of people may decide that you have too much influence, and make it their mission to silence you or discredit you. We have seen this in numerous high profile cases but also in constant reports from women and queer writers, activists and organisers. This is called trolling.

A troll was once just a mountain-dwelling monster in kids stories. Then a troll became the word for early internet users who intentionally sowed discord on IRC and chat forums, often targeting and singling out new users. But now the word is used more broadly to describe people who target and harass others online. This can include anything from constant derogatory and belittling messages to edited images and even threats. Most often the subjects of this kind of abuse are marginalised groups like women, queers and people of colour. In recent years there have been more and more cases of people speaking out about how they are harassed.

Case Study: The Zero Trollerance six-step self-help program!

Hate has always been a part of the Internet and the intentional harassment of other people (termed trolling) has too. But the gendered forms of harassment and violence on Twitter today point to a deeper problem in society that cannot be solved by technical solutions alone. Trolls need serious, practical help to overcome their sexism, deal with their anger issues and change their behavior. Zero Trollerance is a satiric feminist initiative for a self-help journey designed by fake guru Adler King in consultation with reformed trolls. Adler's team of Troll Coaches are constantly analysing Twitter and enrolling new trolls in the program where they are led through an intensive process of self excavation and given practical tools to overcome their inner hurdles. For trolls, this is the first step towards a new life.

Dealing with trolls

There are two key ways you can deal with trolls: one is to block them and report them to the platform you are using or to engage with them. This decision depends on what you want to achieve. Blocking them can definitely work and you can continue with your work unimpeded. A project like “Block Together” was developed to help people who are harassed share their blocklists with each other. Historically platforms like Twitter and Facebook have not handled reports of intimidation and violence very well. However this is changing, as they recognise how severe this problem is and how it deters people with important voices from using their services. When blocking doesn't help is when users are really committed to trolling and create numerous different profiles (called “sock puppet accounts”) to continue the harassment. Then your blocking has to keep up with their new account generation and it becomes tedious. You might consider the alternative of engaging trolls. There are a few tactics for engaging trolls. One is to try and enter into rational arguments with them and interrogate their views. Another way is to try to shame them or use humor to deflate them.

Effective engagement with trolls can actually help to generate a debate and public interest around the act of harassment and involve others online in talking about safe spaces, violence, sexism and online behavior. It can also be a source of empowerment for the subjects of trolling: seeing others laugh at your harasser can be very uplifting.

Swarm and Identity online

The method of swarming can be used to drown out harassers. This can be done in retro style by creating communities of support with your allies in social media spaces where you are likely to encounter harassment. When someone is being targeted, others can quickly be alerted and bombard the harasser with messages. The content of that message is up to you: it could be scolding, educational, or loving. Another option is: instead of directing messages towards the harasser, the swarm can fill the victim's content stream in order to quickly make the negative, violent content disappear into online history.

If you want to engage with trolls, and even if you consider using the “swarming” method, you might prefer to stay anonymous to avoid having your real identity trolled. Setting up a network of second accounts to do your troll-response work can be a good idea for your organisation or your community of friends. It might be easier too, psychologically, to say some of the things you want to trolls, than you would when it is linked to your main identity. And it is more performative: you can create any kind of identity you want and style it with an avatar, a funny name, a character etc (see Chapter XXXX).


While battling the trolls in the old-fashioned human way can be fun and eye-opening, it can also be a time waster. Another option to consider is automation. For this you need someone who can do some coding to start from scratch or work with freely available code someone already uploaded on software repositories such as Github.

A bot is a software application that runs an automated task over the internet. Bots perfom tasks at a greater velocity than humans can. There are many different breeds of bots: for example the spambot which harvests email addresses and contact information or the attention bot which fakes clicks on Youtube videos to make them look more popular than they actually are. They can post content, gather information and click on things. Twitter is filled with bots which use algorithms to harvest information and tweet. Many of these are humorous and random: like @twoheadlines which randomly grabs news headlines and combines them to create funny combinations. The below steps address Twitter mainly. However some of these ideas can be used across other platforms too.

Bot Versus Troll

A bot can be programed to document trolls' activities or talk to them, so that you don't have to. There are a few ways of doing this: the autotweet bot and a silent data-gathering bot in combination with the talking bot(s). The examples below speak specifically about what is possible on Twitter. However the ideas could be applied to other platforms such as IRC chats, forums and other social networking platforms.

1. The data-gathering bot: The data-gathering bot quietly scans Twitter gathering tweets, usernames and any other available information you program it to, and places this in a .csv file for you to analyse or use for further purposes outlined below. This first kind of bot can be useful just for understanding what kind of content is out there and maybe doing a first stage analysis of abuse.

2. The simple tweeting bot: If you follow the #gamergate hashtag on Twitter, you will see a bot called @everyethics which tweets different humorous reasons for the #gamergate trolling, ridiculing of the claim that Gamergate was not about attacking women in gaming but about “ethics in game journalism”. While this bot could be seen as spam, it was actually clearly a strategy to undermine and make fun of the trolls.

3. The retweet bot: The retweet bot is programmed to scan the Twitter API for a list of words, phrases or hashtags defined by you, and to retweet those. This would be a strategy to document and publicise Twitter abuse. Here's an example of such a bot you can download and install.

4. The autotweet bot: The “autotweet” bot is similar except that every time it finds a tweet with one of the words, phrases or hashtags you have programmed it to look for, it will tweet a prewritten tweet to that user. There are a number of examples of this in Twitter history: @stealthmountain which corrects any Twitter user who spells “sneak peek” wrong. These bots get shut down much quicker now as was shown by @fembot which responded to racists and sexist tweets that it spotted and was blocked after making only 75 tweets. Unfortunately Twitter does not make it so easy to do this anymore.

5. The tweeting bot in combination with the data-gathering bot: You can use a data-gathering bot to find the users tweeting violent things, compile them in a spreadsheet for you to read over and check for accuracy and remove any false positives. Alongside the data-gathering bot, you can have a talking bot or a team of talking bots which can tweet whatever you decide is useful information, to those users.

Watch out for

1. Language is slippery: If you want to tackle violence against women online, you will have to be very careful about what kind of language you search for. For example, every time someone uses the word “bitch” on Twitter to intimidate or harass a woman, there are probably at least five other people using it to tell their friend how much they love them or talk about the latest celebrity affair. The best way to figure out which language is used to harm women is to crowdsource it from people who have been harassed and then do a number of tests, pulling tweets from the Twitter API and then analysing it yourself.

2. Twitter is smart (and strict): Twitter is not against bots and if you just want to create a bot that scans information from Twitter for you to analyse, or a bot that just tweets out to no one in particular, you will not encounter any problems. However if you want to tweet @ other Twitter users, you have to take into account Twitter's policy against spam. See Twitter's guide to Automation Rules and Best Practices.

Evading Twitter's spam filters

There are a few things to keep in mind when trying to bypass Twitter's spam filters:

1. Safety in numbers: The more bots you have to distribute the work amongst, the more successful you will be. The group Peng! Collective did such an action in 2015 which they called “Zero Trollerance”. They ran a silent data-gathering bot to identify trolls through a long list of keyword combinations, hashtags and phrases. They then ran 160 bots which tweeted at the database of 3000 trolls, sending them new messages daily for an entire week. They were for the most part able to avoid being blocked by Twitter or users because they had so many bots and they rotated the tweeting across the bots.

2. Rate Limit: Twitter monitors each account's activity and has a “rate limit” that limits your number of tweets, to ensure that no one floods the content stream. This is also the way that they figure out which accounts might be spambots. If the frequency of tweeting looks like it doesn't come from a human, Twitter will block the account. If you try to tweet the same tweet, many times right after each other, you will receive a message telling you that your tweet looks like spam and that Twitter is blocking you in order to protect their users. At the time of writing, tests were done with tweeting 15 times with 8 minute intervals and this passed under the radar.

3. Content: If you are tweeting the exact same tweet over and over again, this is also a red flag to Twitter's spam filters. How to avoid this is to pad every tweet with a random word from a readable language that is not the same as the language you are tweeting in. The easiest way to do this is a compile a long list of these words, and program your script to draw from this list randomly for every tweet.

4. Location of the tweets: Use a VPN which gives you a new IP address every time you reconnect to fool Twitter into thinking that the accounts are being managed from different locations. To learn more about what a VPN is, read "Anonymising your connections" in Step 0.

Setting up a Twitter bot account

You still need to write the script for you or to configure a script already created by someone else and downloaded from Github, or to find someone else to do this for you. But what is easy to do and what even volunteers might like to do is to help you set up all the Twitter accounts so that they can be easily controlled by the script.

1. Create a new account as you would normally and make sure to give it a photo, follow some people and do some tweeting (recently registered, faceless accounts with 0 followers will get blocked very quickly). 2. In order to function as a bot the account needs to be verified with a valid phone number. To do this you can use your own phone number, volunteers numbers or buy a bulk of cheap sim cards. Don't use the same phone number for numerous accounts – again this will be a quick sign to Twitter that the account is dodgy. 3. Now you need to register an application with the Twitter API which will allow your bot to make “calls” to the API, i.e. retrieve or send data. Go to apps and create a new application. You can provide any dummy content in the fields there and then you can set your permissions to “read and write” and generate the keys you will need.

More on these steps and some simple bots to download and test out at Cyber Guerilla.