Tagging difficult users with infectious markers...
October 28, 2003
Following on from my earlier piece on Stealth Moderation I thought I'd talk a bit about a technique we've been using on Barbelith recently to deal with a particularly thorough and unpleasant troll-attack. But first I should recap on the specific situation that we're trying to resolve with this technique.
One of the great difficulties with looking after an online community is that it's generally almost impossible to ban a user from a site if they're dedicated to breaking in. The only circumstances you can ban them are when you require payments via credit cards, hard-to-obtain unique forms of real-life identification or when you're prepared to take the situation to the police. Otherwise all they have to do is sign up for a free e-mail account, and re-register on your site. Within ten minutes they can be back causing trouble, your ability to set the rules for your community space has been completely undermined and there's very little you can do about it.
And that's only one use of multiple user names. Many trolling users will maintain several concurrent accounts, which they will use to support the position of their prime identity - making all online battles seem larger and more significant than they actually are and obfuscating the fact that - at heart - it's just one troublemaker working quite hard to spoil the experience for all the others. These alternative user names are often known as "Sock-puppets" for vaguely obvious reasons. Typically a troll of this kind will use their sock-puppets to post self-supporting messages like, "Hey, why are you being so down on the guy. I think he has a point and you're all being really **** about it". I've seen people using these multiple user names to create identities that are almost identical to other user's self-representations (a duplicated character in the username - or sometimes just a space after their name, depending on the software) and then using that identity to suggest that their alternative usernames, "might have a point - maybe it's best not to wind them up any more", or even to suggest that their alternative trolling identity might have started investigating legal recourses. Even stopping new registrations won't necessarily stop this kind of activity as long as the e-mail addresses of long-dormant users are available to be contacted and appealed to. And there will always be one user who has two or more user names who believes any kind of ban is a de facto attack and will support a long-term troll, however obviously destructive (or even illegal) they might be...
Essentially it all boils down to one problem - that you can ban user names easily, but it's far from easy to ban real-life people. There are many approaches to this kind of problem, but one thing is clear - on occasion users do need to be banned - however much we may wish it otherwise.
One approach that we've been using recently with a fair amount of success (although it breaks my first and most important rules of what constitutes a long-term successful moderation strategy) is based around finding ways of demonstrating clear links between user-names - links that indicate that they are being used by the same real-life users or groups of users. We used cookies again, so it's only going to work on platforms where you are using either a web-based interface or write the client-side software, but it really has proven extremely useful.
A user who we wish to tag is marked as tagged in the user table of the database. When they next login, a cookie is placed upon the browser that they use. From that moment on, any other user-name that logs in via that machine will immediately and automatically be tagged in turn. If that latter user then moves to a different computer and logs in, that computer too will have a cookie on it that marks it as being 'used by trolling users' - and any subsequent logins on that computer by different user names will result in those user names also being tagged. At the individual level this can mean that each new user name can be directly and quickly identified as belonging to a troublesome user, but it gets even more useful when a group of users decide to share a new user name to cause trouble on a board. Everyone of them will be tagged next time they login.
In order to make the process more useful, you can find ways of adding more information to the cookie. One particularly useful piece of information is which tagged user-name triggered the site to leave a cookie on someone's computer. This information can be particularly useful if you're unlucky enough to have attracted the attention of semi-organised groups of long-term troublemakers, since it allows you to track the course of your tag through the community and - in turn - enables you to clearly see specific relationships between individuals.
What you choose to do with this data is another matter entirely. In order to avoid many of the fairly obvious ethical issues that surround tracking user information at this kind of level, we've operated on the basis of revealing to the user that they have been banned, placing the cookie immediately on their browser and then waiting for them to try other usernames which in turn will then automatically and immediately be banned. Obviously this approach is not without its problems - for a start it makes it easier to determine what is causing the bans (particularly for the more technically literate) and may help a dedicated long-term troll find workarounds - so you might want to obscure the issue a bit by triggering a user name ban after a random number of hours or posts so there is this perception of human agency behind the scenes. Either way, it's probably best not to name the cookie after the banning process, as that might give the game away...