Moderation and Censorship

I've noticed a spike lately in blog postings about Facebook's "censorship" of users. It's a perennial issue, and Facebook isn't the only company to come under fire of bloggers, for whom freedom of expression is sacrosanct -until the moment that someone begins spamming their own blog. Funny how that works.

But more to the point, there is not yet a foolproof system of moderation for online forums and communities. Though some progress has been made in that area, the problem is inherently difficult to address - and while I don't have a good solution to share, I wanted to capture some thoughts about the solutions that have been tried:

1. No Moderation

The Internet began as an open system, without any moderation whatsoever. Even before the Web, there were bulletin boards (USENET) and chat forums (IRC) where anyone could show up and post anything they liked.

That didn't go over so well: unscrupulous advertisers, juvenile pranksters and attention-seekers, and those whose motives were to abuse the system to choke off any conversation that didn't suit their liking flooded the boards with posting that the vast majority of other users found annoying. It wasn't long before USENET began to provide moderated threads and IRC created features to set up private forums and enable room creators to boot users who misbehaved.

I don't think anyone could successfully argue that having no moderation at all is a viable solution. It's an attractive theory, but has never, to my knowledge, been effective in practice for any forum with a substantial membership, as there are always a few miscreants in the woodwork who simply have to be dealt with if the forum is to serve its intended purpose for the majority of users.

The greater the body of users, the larger the number of miscreants (and the greater its attractiveness to them), until the latter completely corrupt and pollute the conversation threads, rendering them useless. I've not seen statistics on this, but from my own experience, I'd guess that the line is drawn at about 10,000 users (though sometimes much sooner, depending on the forum and its topic).

2. Individual Filtering and Blocking

The next step up from no moderation at all is providing the user with the ability to filter and block. To disambiguate: "filtering" is hiding messages based on their content and "blocking" is hiding messages based on the identity presented by the sender. (And to escape clutter, the single term "filtering" will be used hereafter).

This solution predates moderation, in that IRC users could "mute" an obnoxious participant and USENET users could "killfile" posts. It's a democratic approach that allows each individual to decide what information to bother reading - and if there was any problem with the filters hiding desired messages or failing to hide undesired ones, the user had only himself to blame.

The drawbacks to this approach are inefficiency and ineffectiveness. Each user must attend to the task of filtering. Since those who wish to abuse a forum repeatedly are elusive: they may change their login ID or alter the content of their messages to get past the filters, so the task of filtering, and they will do both repeatedly, making effective filtering an ongoing and tedious task that, ultimately, is not very accurate.

Moderation by a central source is, in this way, merely the consolidation of blocking and filtering, assigning the task to a moderator who has the interest, time, and expertise to clean up the detritus for the benefit of the community. But naturally, when a task is delegated to another person, not everyone is pleased with they way they go about doing it. This is at the heart of the problem, and it is this very quality that continues to make it difficult to solve.

3. Operator Moderation

The first form of moderation is "operator" moderation - by which I mean that a human being, either the owner of the forum or appointed by the owner of the forum, tends to the task of filtering content for the participants.

This form of moderation relies on human judgment, which is both its strength (in that human beings can more accurately identify information that doesn't fit the purpose or rules of a forum) and weakness (in that not all humans agree, that human judgment is subject to error, it may be applied inconsistently, and humans have a limited capacity to process information).

Moreover, objections to moderation are at their most vehement when a human moderator is involved: accusations of prejudice, intention to suppress, etc. are given more credence when a person is making the decisions as to what others are "allowed" to post in a forum. And, in fact, moderation is a form of censorship, as it involves an estimation (or decision) of whether a given message is something the community wants to see or the owner wishes to have published on his forum.

If it weren't for the litigious nature of contemporary society, forum operators might find this option more attractive - but human moderation puts the operator in a precarious position whenever someone whose post has been blocked cries "foul" and threatens a civil rights lawsuit, or when others who object to a given post threaten to alert the authorities about the kind of information the operator is "allowing" to be published on the forum he operates.

4. Automated Moderation

To be precise, automated moderation is computer-assisted moderation: software is used to detect and filter unwanted messages and, when a given user has sent more than his quota of nuisance messages, to block or disable the account.

But automated moderation brings problems of its won: chiefly, that computers are very faithful, but not very smart at all. A system can be automated to block abusive messages ("John is an ass") but will erroneously block inoffensive ones ("John is an assistant principal" or even "No one should say that John is an ass"), and due to the complexities of language, the limitations of computer logic, and the inventiveness of miscreants, no reliable solution has been found.

Chiefly, this is because computers treat language as a jumble of letters and are unable to derive their meaning. While technology and techniques have advanced beyond mere pattern-matching in text to attempt to derive a sense of the ideas the text conveys, it remains even more incompetent at that task than it is at the simplest forms of pattern matching.

Also, automated moderation does not escape the problem of operator bias. While software can be consistent and objective in applying the criteria for filtering content, a human being must program and/or configure the logic that is applied by the system, so it (rightly) does not avoid accusations of bias. If anything, it makes bias more efficient and, by virtue of its own glitches and limitations, makes that bias seem more arbitrary and draconian.

5. Crowdsourcing Moderation

A recent fad in moderation is "crowdsourcing," which combines the various techniques listed above: there is no initial attempt at moderation, then individual users set their filtering preferences (of flag specific messages), and the system attempts to amalgamate these individual preferences into an automated filter.

This would seem to be an evolutionary leap, combining the best qualities of all previous forms of moderation, but it often combines all of the worst qualities as well: it requires redundant efforts of each user, depends on the users to be objective and fair in their filtering criteria, and uses their input to generate automated rules that may cause innocuous content to be blocked.

In addition to these problems, crowdsourcing moderation creates a problem of its own: ganging. A group of individuals can band together to take control of the forum by collectively seeking to control the topic by blocking the same posts from the same individuals, such that these participants can control the content available to the majority. This "ganging" effect may be the result of an intentional conspiracy of individuals to silence a particular person or point of view, or it may also be an unintentional side effect of the same individuals acting independently, without a conscious decision to gang together.

Arguably, that's merely a matter of fine-tuning to manage the balance, such that a small number of users can't skew the filtering algorithm, but it doesn't require so many users to agree that it undermines the efficiency of crowdsourcing. But the same argument has been made of the other forms of moderation, and after decades of effort, the problem has yet to be solved.

***

In the end, this post has been just another examination of the problems from a guy who doesn't have any solutions, and is seeking to better understand the phenomenon - but that's largely the point: some problems are inherently difficult, and there is no simple solution, though it seems that having a better understanding of them makes it a bit easier to swallow, and provides a bit better defense against those who constantly seem to be attempting to spin up panic or outrage at a situation that, in spite of the best efforts of many people, has yet to be effectively addressed.