Mod 201: Community Self-Moderation

On Saturday, March 5 the first version of Community Self-Moderation Tools (mod tools) was launched. This blog post is intended to 1) describe the history of b@b moderation/trashing, 2) to describe the behavior and functionality of the new system and 3) explain the next steps.

I.  History

Previous iterations of b@b attempted to solve moderation issues with a "trash" feature. The trash feature would remove a specific post after a certain number of users had marked it as such. This would remove the post from view for everyone. This feature had gone though a number of tweaks and major versions over the years:

• Version 1 (2006-2008) - A post was removed after X number of users voted "trash" from different IP addresses (say, 5 votes). This model broke down when it was discovered that rotating IP addresses allowed any one individual to remove posts with enough determination. Some users would run around campus finding different computers to trash the posts that they wanted eliminated. Some users also figured out that, in some circumstances, logging in and out a computer would give you a new IP address allowing you to vote or trash infinitely. In some cases, users were going crazy running around campus trying to remove posts that were personally offensive. In other cases, legitimate posts were being removed based on a personal agendas.
• Version 2 (2009-2010) - This version implemented a rather extensive algorithm to detect specific types of words or phrases and weighted them based on an "objectionability index". In theory, this was the end-all-be-all of preventing users from breaking the rules. The b@b admin could set a list of predefined "bad words" and "horrible words". These lists would include words like bitch, cunt, fag, nigger, etc. and the system could identify any perversion of these words. For example, the lemmatization of "bitch" "bbbbbitch" "b1tch" "b|tch" "bbbbbbb*7ch" would all be flagged as the same meaning and would be given a high objection index. Also, the frequency of bad or horrible words in a single post would increase its overall index. Ultimately, a "bad" post would sometimes require only 1 or 2 votes to have it removed completely while others would require a significant number of votes. The problem with this system was that 1) it often accidentally stifled free speech (I can make arguments that contain the word bitch but in a constructive manner) and 2) it couldn't account for the fact that name calling and trolling could still fly under the radar.

Bottom line, free speech is of paramount importance. Automated systems aren't particularly good at handling posts subjectively.

Realizing this, it was decided to fall back on something more rudimentary with the introduction of the "report" feature. The idea is to allow any user to report any post for violating one of the rules. Then, a sys admin or moderator could sift through a list of reported posts and decide an action (warn, dock points, block temporary or permanently ban). This system has been surprisingly successful relative to the level of distress b@b has historically caused the community. As the service has grown, however, there has been scalability issues. It is difficult for any one moderator to moderator the site 24/7.

So let's try something new.

II.  Community Self-Moderation v1

B@b now awards certain users "moderator" privileges based on the user's seniority and the number of users online at any given time.

What can mods do? 

Moderators are asked to use their best judgement to "warn" users based on specific definitions of tolling, libel, hate-speech and spam. If a post is found in violation, a moderator can:

1. Just warn - The user instantly gets a notification that they've been warned and told the reason why.
2. Warn and dock 50 points - There user is warned and notified, but also docked 50 points.
3. Warn and dock 100 points
4. Warn and dock 200 points
5. Warn, dock 200 points, and temporarily ban the user from the site. Temporary bans are either 15 minutes or 30 minutes depending on the type of offense. This is reserved for repeat offenders and should only be used in the most extreme cases.

Who can be a mod? 

Moderators are designated base on seniority (age of account) and points acquired. The number of moderators on the site at any given time is variable depending on the number of users online. For instance, there needs to be roughly 10 users online before any 1 user is given moderator privileges. In most scenarios there are between 1 and 2 moderators, sometimes 3. The maximum number of moderators (no matter how many users are online) is 5. You are moderator if you see [mod] next to your "Account" button in the header. Mod privileges are dynamic: if someone more senior logs in, the less senior user will forfeit their mod privilege. (For those of you confused to see [mod] appearing and disappearing, this is why: permissions are being shuffled based on who is online.)

Wait, can't a mod abuse their power? 

This was carefully considered before implementation. First, moderators can only warn or take action against an offender a maximum of 5 times per hour (this is subject to change based on experimentation). If you are a moderator who has gained and lost privileges quickly, your number of warns per hour still holds (it does not reset). Finally, this needs to be perfectly clear: I monitor all moderator transactions in near real time. I am watching carefully to be certain the system is not being abused. If I see an obvious abuse of privileges, I reserve the right to dock points, block or delete the user permanently from the site. The community should take comfort that this is not a free-for-all system. If I discover any account attempting to "troll for points" in an attempt to rise to the top, I will not hesitate to deactivate this person permanently.

III. Next Steps

So far, I've been pleasantly surprised at the results of the first few days since implementation. There has been no instances of abuse of privileges and all users warned have been completely legitimate (I would have done the same). I understand that this is not a perfect system so I will keep any eye out and continue to ask for community advice on how it can be improved.

I believe the next step will be to give moderators access to a master list of reported posts. This way, moderators will be able to extend their reach to clean up posts that their peers have reported kindly. The system would be designed so that a moderator could only remove posts that were reported by other users (a mod couldn't report and remove a post on their own). This seems to be the next logical step, though, it is still a half-baked plan. Your ideas and feedback would be greatly appreciated.

With that said, I hope that this frames the new feature set adequately and clears up most of the common questions. As always, if you would like to feedback on the site please feel free to do so, send me a message via the "Make a Suggestion" box or shoot me an email.

Jae