The new campaign to ‘Reclaim the Internet‘, to ‘take a stand against online abuse’ was launched yesterday – and it could be a really important campaign. The scale and nature of abuse online is appalling – and it is good to see that the campaign does not focus on just one kind of abuse, instead talking about ‘misogyny, sexism, racism, homophobia, transphobia’ and more.
There is more than anecdotal evidence of this abuse – even if the methodology and conclusions from the particular Demos survey used at the launch has been subject to significant criticism: Dr Claire Hardaker of Lancaster University’s forensic dissection is well worth a read – and it is really important not to try to suggest that this kind of abuse is not hideous and should not be taken seriously. It should – but great care needs to be taken and the risks attached to many of the potential strategies to ‘reclaim the internet’ are very high indeed. Many of them would have precisely the wrong effect: silencing exactly those voices that the campaign wishes to have heard.
Surveillance and censorship
Perhaps the biggest risk is that the campaign is used to enable and endorse those twin tools of oppression and control, surveillance and censorship. The idea that we should monitor everything to try to find all those who commit abuse or engage in sexism, misogyny, racism, homophobia and transphobia may seem very attractive – find the trolls, root them out and punish them – but building a surveillance infrastructure and making it seem ‘OK’ is ultimately deeply counterproductive for almost every aspect of freedom. Evidence shows that surveillance chills free speech, discourages people from seeking out information, associating and assembling with people and more – as well as enabling discrimination and exacerbating power differences. Surveillance helps the powerful to oppress the weak – so should be avoided except in the worst of situations. Any ‘solutions’ to online abuse that are based around an increase in surveillance need a thorough rethink.
Censorship is the other side of the coin – but works with surveillance to let the powerful control the weak. Again, huge care is needed to make sure that attempts to ‘reclaim’ the internet don’t become tools to enforce orthodoxy and silence voices that don’t ‘fit’ the norm. Freedom of speech matters most precisely when that speech might offend and upset – it is easy to give those you like the freedom to say what they want, much harder to give those you disagree with that freedom. It’s a very difficult area – because if we want to reduce the impact of abuse, that must mean restricting abusers’ freedom of speech – but it must be navigated very carefully, and tools not created that allow easy silencing of those who disagree with people rather than those who abuse them.
One particularly important trap not to fall into is that of demanding ‘real names’: it is a common idea that the way to reduce abuse is to prevent people being anonymous online, or to ban the use of pseudonyms. Not only does this not work, but it, again, damages many of those who the idea of ‘reclaiming the internet’ is intended to support. Victims of abuse in the ‘real’ world, people who are being stalked or victimised, whistleblowers and so forth need pseudonyms in order to protect themselves from their abusers, stalkers, enemies and so on. Force ‘real names’ on people, and you put those people at risk. Many will simply not engage – chilled by the demand for real names and the fear of being revealed. That’s even without engaging with the huge issue of the right to define your own name – and the joy of playing with identity, which for some people is one of the great pleasures of the internet, from parodies to fantasies. Real names are another way that the powerful can exert their power on the weak – it is no surprise that the Chinese government are one of the most ardent supporters of the idea of forcing real names on the internet. Any ‘solution’ to reclaiming the internet that demands or requires real names should be fiercely opposed.
Algorithms and errors
Another key mistake to be avoided is over-reliance on algorithmic analysis – particularly of content of social media posts. This is one of the areas that the Demos survey lets itself down – it makes assumptions about the ability of algorithms to understand language. As Dr Claire Hardaker puts it:
“Face an algorithm with messy features like sarcasm, threats, allusions, in-jokes, novel metaphors, clever wordplay, typographical errors, slang, mock impoliteness, and so on, and it will invariably make mistakes. Even supposedly cut-and-dried tasks such as tagging a word for its meaning can fox a computer. If I tell you that “this is light” whilst pointing to the sun you’re going to understand something very different than if I say “this is light” whilst picking up an empty bag. Programming that kind of distinction into a software is nightmarish.”
This kind of error is bad enough in a survey – but some of the possible routes to ‘reclaiming the internet’ include using this kind of analysis to identify offending social media comments, or even to automatically block or censor social media comments. Indeed, much internet filtering works that way – one of the posts on this blog which was commenting on ‘porn blocking’ was blocked by a filter as it had words relating to pornography in it a number of times. Again, reliance on algorithmic ‘solutions’ to reclaiming the internet is very dangerous – and could end up stifling conversations, reducing freedom of speech and much more.
Who’s trolling who? Double-edged swords…
One of the other major problems with dealing with ‘trolls’ (the quotation marks are entirely intentional) is that in practice it can be very hard to identify them. Indeed, in conflicts on the internet it is common for both sides to believe that the other side is the one doing the abuse, the other side are the ‘trolls’, and they themselves are the victims who need protecting. Anyone who observes even the most one-sided of disputes should be able to see this – from GamerGate to some of the conflicts over transphobia. Not that many who others would consider to be ‘trolls’ would consider themselves to be trolls.
The tragic case of Brenda Leyland should give everyone pause for thought. She was described and ‘outed’ as a ‘McCann troll’ – she tweeted as @Sweepyface and campaigned, as she saw it, for justice for Madeleine McCann, blaming Madeleine’s parents for her death. Sky News reporter Martin Brunt doorstepped her, and days later she was found dead, having committed suicide. Was she a ‘troll’? Was the media response to her appropriate, proportionate, or positive? These are not easy questions – because this isn’t an easy subject.
Further, one of the best defences of a ‘troll’ is to accuse the person they’re trolling of being a troll – and that is something that should be remembered whatever the tools you introduce to help reduce abuse online. Those tools are double-edged swords. Bring in quick and easy ways to report abuse – things like immediate blocking of social media accounts when those accounts are accused of being abusive – and you will find those tools being used by the trolls themselves against their victims. ‘Flame wars’ have existed pretty much since the beginning of the internet – any tools you create ‘against’ abuse will be used as weapons in flame wars in the future.
No quick fixes and no silver bullets
That should remind us of the biggest point here. There are no quick fixes to this kind of problem. No silver bullets that will slay the werewolves, or magic wands that will make everything OK. Technology often encourages the feeling that if only we created this one new tool, we could solve everything. In practice, it’s almost never the case – and in relation to online abuse this is particularly true.
Some people will suggest that it’s already easy. ‘All you have to do is block your abuser’ is all very well, but if you get 100 new abusive messages every minute you’ll spend your whole time blocking. Some will say that the solution is just not to feed the trolls – but many trolls don’t need any feeding at all. Others may suggest that people are just whining – none of this really hurts you, it’s just words – but that’s not true either. Words do hurt – and most of those suggesting this haven’t been subject to the kind of abuse that happens to others. What’s more, the chilling effect of abuse is real – if you get attacked every time you go online, why on earth would you want to stay online?
The problem is real, and needs careful thought and time to address. The traps involved in addressing it – and I’ve mentioned only a few of them here – are also real, and need to be avoided and considered very carefully. There really are no quick fixes – and it is really important not to raise false hopes that it can all be solved quickly and easily. That false hope may be the biggest trap of all.
This post originally appeared on Paul Bernal’s Blog and is reproduced with permission and thanks