Social Software and Sociopaths
Somebody posted a picture of “himself” on Moblog this morning. It was bound to happen eventually. Like many people who run communities (myself included), Mat had hoped it wouldn’t happen and was kind of unprepared for it when it did. Palmpixel flamed the poster and asserted that his post would be deleted and it eventually was, when Mat was woken by Norm.
Earlier in the day, Ladislav linked to Julian Dibbell’s A Rape In Cyberspace, an account of a rather nasty incident of a similar tenor which occurred on LambdaMOO back in the mists of time. I’ve just read the article. The eventual arrival of sociopaths in any social system seems inevitable. What their causes or actions are is fairly irrelevant really, the tricky question is how to identify them and protect the society from their actions.
In LambdaMOO the player was killed by a wizard and a system of voting for wizardly enactments established, on Moblog the post was deleted by Mat and several other core users were given mod rights (these include me, though not a core user, I am a good guy). The user continues to have his account, though his crime was rather less severe than the LambdaMOO rapist’s. Interesting that the two solutions were so similar, different really only in scope. The question is highly political and perhaps the most burning conceptual issue in discussion in the area. As Dibell discusses, your view on this issue is completely tied to your political outlook - anarchist, libertarian, conservative, etc. But this doesn’t help anyone design software.
The easy answer for a developer is the same for a Roman general presiding over a Jewish city horribly offended by the actions of one of their own: wash your hands. Give the group a mechanism to punish or banish or otherwise isolate the offender. The core question, as far as I can see, is what type of model you choose for leveraging that power: democracy (true or representative), oligarchy, even autocracy. How that is implemented then becomes comparitively straightforward and is necessarily tailored to the style of your community and the manner in which your users interact with one another.
So the reason I discuss this is really to look at what solution I might find for Hypothetical. I think the key is isolating the core group. On Moblog, Mat has decided to do this manually, for now at least. Being able to learn from his mistakes and those of others, I want to build in an automatic system from the start. Besides group owners already have the ability to kick people (though not permanantly as yet). There are several useful ways to establish core group membership, some examples of which are:
- Total post count
- Post frequency
- Read frequency - based on page hits or something
- Account age
The best solution, I guess, would be combine each of these factors, weighted, into a composite metric.
After that, the next decision is how big should the core be? What is the cutoff? The easiest way to do it is to say I want x members and just take the highest ranking users. Another mechanism might be to identify a baseline level of activity and say that anybody above that can be a core member. The vital issue is that whatever cut off is used, it excludes the sociopaths. Once they become core group members, the system will become confused. Democracy within the core group could be employed to protect against this situation, but I think that would rather inefficient in Hype’s case, and the core would fail to act quickly to remove offensive content before the rest of the group are exposed to it, negating the usefulness of the core group system. Bear in mind that the definition of what constitutes a sociopath will be specific to each group. Penis-posting and virtual rape may be the main activity for some groups and so it would be wrong to hardcode prevention of these things.
<edit>Or I could just let group owners appoint the mods for a group. Doh. I have to start thinking more like somebody who’s providing tools to clever people who will face the same problems that all the wizards on the net have faced at some point, rather than just thinking of myself as the lone Hype wizard.</edit>
What powers to give the core is pretty simple: ban users, remove posts, remove files. There is a complication in the case of banning users, they can just get a new account. I first thought that a pure democratic system would be good. If five people think a post is bad, it’s gone. But then you just create five accounts and you’re a god. The freeness of accounts is always going to be a thorn in the side of Hypothetical, whilst simultaneously being essential to it’s functioning. The same was true of LambdaMOO, as the rapist returned with a new account days later.