Brian Leiter, a University of Texas law professor, does a nice line in mocking opponents and imperially deciding that one commentator wiped the floor with another, or that another law student is a "naif." So it's amusing to see him wax authoritative about a subject that has been thoroughly explored by others and is outside his area of expertise. It's even better to see Heidi Bond take him to task for it.

The debate itself concerns Xoxohth, an online discussion board for law students that Carey finds useful and Prof. Volokh thinks sounds like an old D&D character of his. Oh, yes, and among it's other nicer features, there's a lot of racist and anti-semetic comments raised by the kind of folks who have existed since well before the dawn of USENET. (UPDATE: Since someone wrote asking me to state the obvious, the offensive comments are not one of the "nicer" features, they're interspersed amongst other features that can easily qualify as "nice." The offensive comments are mostly dumb. I must struggle more diligently to avoid ambiguity in language.) What to do about them has been an ongoing debate. Leiter states:

Put aside ethical obligations, and let's just consider good taste and decency: how hard can it be for Messrs. Ciolli and Cohen to delete all the threads with certain words? And if they did that a few times, no doubt the infantile morons responsible for most of this garbage would give up and go elsewhere.

(emphasis added) Now, for someone who likes calling people "naifs," this seems uncommonly naif-like. Having moderated a number of bulletin boards and being old enough to remember the early days of USENET, I can't recall this tactic ever having been particularly successful. Deleting a few posts makes disruptive users go away? It's certainly not worked at 3YoH, though I try to keep the community here relatively friendly. Perhaps Prof. Leiter associates with a better class of disruptive user.

Heidi points out the problem quite nicely:

I also think that Professor Leiter is entirely naive about the effectiveness of filters. Just try filtering out the word "fag". Next thing you know, the perpetrators post "f@g". And then you filter that, and they start saying "ffag" or "fa-g" or "f.ag." It is easier for them to generate patterns that imply "fag" than it is for the few humans in charge of the site to generate filters. Furthermore, the racist, sexist, and abusive crap that comes out is not necessarily a result of language. How do you mechanically filter out a racist discussion bashing Blacks for racist reasons without also mechanically filtering out the discussions on affirmative action?

Indeed, trying to figure out an effective method of filtering comments without losing posts that are important has been one of the holy grails of the internet for as long as I've been using it. No one's managed it yet. Consider, for instance, the MT-Blacklist program that I use to stop comment spam. Every so often, I find that a perfectly innocent comment has been blocked because I set the filter with an overinclusive term. (I've caught this twice, and emailed an apology to the commentor, but doubtlessly I miss more than that.) The false-positive rate on such a filter isn't very high, because the MT-Blacklist algorithm is mostly fighting other machine-based algorithms, and a sufficiently-determined human can get around my blocking. What Leiter proposes is much harder: a machine that is trying to stop an individual human determined to write a single post.

The other option is human moderation. In an update, Leiter mentions Nontradlaw, which apparently fully moderates its forum. Moderation is a tradeoff, a high-cost strategy because human intervention is continually required, and thus a limiting factor on the size of the system. There are ways of reducing this--the user-moderation systems at Slashdot or DailyKos, for example. These are effective but complex systems, and again raise the price for the system administrator. [1] (Xoxoxth does have an "on/off topic" flag, but it doesn't appear as complex a system as Slashdot's.)

I actually disagree with Heidi on the propriety of deleting comments, as she's much more reluctant to do so than I. If someone leaves a racist or homophobic comment--or even one that I just feel is needlessly offensive--I may very well delete it. Then again, I've left up comments like that simply so that someone else could trash the commentor. (Or in one case, I went ahead and published the name of the firm at which one commentor worked, based on his IP address--I've made my feelings about anonymity well-known.) But that's because 3YoH is very much my project and associated with me: it's a blog, not a bulletin board. Not saying that Heidi's policy is wrong: it's merely hers, as mine is mine.

But what if I missed some horrible comment? Am I under some duty such that I've then become negligent? Have I been ethically lax in providing a forum for a nutcase? Or can we take it as read that those who run forums may choose to expend resources on other things than moderation, and that the occasional troll or flamer doesn't invalidate the usefulness of the board?

In any event, here's a challenge for Professor Leiter. He's got a high-traffic blog with commenting capability. He apparently thinks that deleting the comments, either through filters or manually, is a relatively simple task. Why not prove it? He can open up comments on all his entries, and I'm sure the good folks at Xoxoxth who he's been condemning would be happy to provide a number of comments for him to weed out. If he really thinks that creating this holy grail will be so simple, let's let him write the code. After all, if the man can drive flamers to the point of extinction, I'm sure he can more than triple his salary with the profits he makes from his new software.

Are you looking forward to LeiterFilters 1.0? I know I am.

[1]: For a good discussion of the costs of implementing a Caio M. S. Pereira Neto, Online Collaborative Media and Political Economy of Information: A Case Study, 21 J. Marshall J. & Info. L. 511 (2003). This study compares and contrasts the Slashdot and Kuro5hin accreditation systems and discusses their different goals and outcomes. More than likely, this is the kind of thing that Xoxohth needs, but implementing it is much more than the trivial task Leiter suggests.