Since HopelessN00b wants to have this discussion in public, I will oblige.
I was on the site today because we got a few different emails this past week about it. One was Chris S, stepping down as a moderator - I'll let him talk about that if he wishes to; moderators are volunteers, and are of course free to step down at any time. The rest were complaints about actions taken by HopelessN00b. This is also not particularly unusual; we get complaints daily about moderators on all sites. Usually I spend a few minutes checking out the situation, a few minutes more explaining it, and everyone goes on with their lives.
Instead, I found this:
(actually I was looking at a different page that displays the last 500 comments, but the gist is the same: a wall of identical comment, posted in groups, each a few seconds to a few minutes apart)
This is a bit worrying, but not entirely unusual - moderators or regular users engaged in a big cleanup campaign often have histories of repetitive actions. So I dug a bit deeper - starting with the links in the comments themselves, then in recent meta activity. Nowhere did I find any hint of a massive question cleanup campaign. Now I'm getting a bit worried, especially since these closures (and comments) are still appearing while I'm researching. I checked chat - both the public rooms associated with Server Fault, and the private moderator room. Nowhere was there any discussion of this, save for a lone message from Hopeless in a nearly-abandoned chatroom noting that he had been "closing webpanel questions by the hundreds".
At this point, I'd already spent an unexpected amount of time on this, and from the look of things I'd be spending a lot more. And closed questions were still piling up; as a final sanity-check, I reviewed a sampling of them - some were obviously questions about administration panels, but this wasn't consistent; the majority of those I checked made no mention of such tools.
##Handling a rogue moderator
When a normal member of the site starts making massive, controversial changes without prior discussion, the standard procedure is to immediately suspend the account in order to stop the bleeding, then discuss the matter. But suspending a moderator doesn't accomplish much. So the remaining option is to remove moderator status, and then attempt to discuss whatever is going on - if the situation can be resolved quickly, this need not be a permanent change in status; if the situation goes south, suspension is then an option.
So I removed Hopeless's moderator status and sent him a message requesting that he explain the situation (the other Server Fault moderators were also copied on this message).
I've yet to receive a response.
Thus far, I've identified 572 questions closed with some variation on that comment over the past 2 months. The average score was 0.1, the maximum score was 19. The comment was replied to 29 times; Hopeless responded to 1 of them. These questions will need to be reviewed; Michael has started a separate discussion that will help determine the criteria which should be applied to them.
##Update 4/10/2015
After being asked to post this publicly, I finally got a response privately last night. It did not attempt to answer the one question I asked.
I'll be going through server logs this weekend to attempt to determine what was actually being done here.
So in summary, if you are a community moderator on a Stack Exchange site, here’s what to expect:
- As a moderator, your actions now represent the community, so you will be held to a higher standard of behavior. You are an ambassador of trust, with the same sorts of rights that the official development team and community coordinators have.
That trust has been betrayed. I am extremely disappointed in how this situation has played out. I will be working directly with the remaining moderators to ensure that this community's standards are being upheld and communicated effectively, and that any damage is repaired.
##Update / conclusion: 4/13/2015
I've finished my analysis of Hopeless's activity as a moderator on Server Fault. This is a bit long; there was a lot of activity - more on that in a bit. For those of you who aren't interested in the details, the summary is that I was relieved to find that Hopeless has acted as an enthusiastic but largely competent moderator, serving the community here faithfully except when it came to responding to questions/concerns regarding his actions and in communicating his efforts in cleaning up old posts. Related to this, I've identified two areas where our tooling is deeply lacking and likely exacerbated the problem.
Following his election in December, Hopeless was off to a good start: he handled a respectable portion of the flags raised here, participated in review, and participated actively on meta. The only oddity I found in his actions at the start was a group of 581 questions that he locked for Historical Significance in his first week; this is notable for two reasons:
- only 241 of them were closed prior to being locked, and
- that's more questions than had previously been locked in the entire history of the site. (for reference, historical locks are a bit special - it was added for "too big to fail" posts that were no longer allowed but too good to lose, and posts where it is applied disappear from the homepage and from normal question lists... So it's a bit hard to notice when it's being used - more on that later.)
A good chunk of these were subjective polls and GTKY stuff from the very early days of the site - stuff like What's your favorite Linux distribution? or What is the best VPN technology to implement in a SOHO setting? it's hard to get upset about locking them, unless you're inclined to complain that they weren't just straight-up deleted. Still, this was a sign of things to come in that I can find no discussion of it anywhere.
Then in early January, things started to get interesting: Hopeless started cleaning up tags. Starting with a set of web admin tags he'd identified back in September, and then moving on to a much larger group of tags. The week of January 19th he edited tags on 875 questions, closed 399, deleted 221, and locked 817; he then slowed down for a few weeks before breaking his own record during the week of February 16th with 988 tag edits, 262 closures, 546 deletions and 723 locks... and 1,513 tag merges.
The locks and merges need a bit of explanation. In the big tag burning thread, he mentions using locking as a tool to prevent bumping too many questions to the homepage:
questions that are locked for historical significance can be ?-edited (therefore, retagged) without bumping them to the top of the active list... so I'll be temporarily locking questions I retag to minimize disruptions.
Not all of the locks were temporary; of the 2342 questions locked in January and February, 869 remain locked.
The merges are harder to explain. Indeed, someone asked about them on February 26th, but didn't recieve any clarification. We discussed this in chat at one point, after I'd noticed the merges and complained; apparently, the idea was to get rid of the tags first by merging everything into [off-topic] and then go back through that tag and perform any necessary cleanup. It helps a bit to read this answer, in which Hopeless responds to a user complaining about a tag he was using being suddenly gone; if you think of tagging more in terms of a physical filing system rather than a folksonomy, the concern over having too many of them is understandable.
January also marks the start of using a canned comment when closing questions, with the week of the 12th seeing a whopping 312 questions closed with:
This question appears to be off-topic because it is about
working with a service provider's management interface, such as cPanel.
By late February, Hopeless had refined this message to its current form, and was using an app to integrate it into the moderation UI. Note that the question I just linked to is the only discussion of this I've been able to find, and concerns a newly-asked question. This becomes easier to understand upon realizing that a substantial portion of these questions were also being locked, at least temporarily - thus they would've immediately disappeared from all normal question lists, and their authors - although still notified of the comment - would be unable to respond to it.
After the initial tag cleanups, Hopeless was relying less and less on tags to filter questions and more on keyword searches for the names of various web administration tools. He was still using [off-topic] as a filter for retagging, but by the week of March 30th retagging had taken a clear backseat to closing: 74 retags, 227 closures (95 of them coupled with the "webadmin" comment), 67 deletions and 74 locks. Last week saw 110 closures (79 webadmin), 82 deletions, 33 locks and 35 tag edits.
To recap: this cleanup started with this meta post, where the only mention of closing is in answers noting that normal rules should be applied. It continued through a tumultuous tag burnination and transitioned into an effort to eradicate all mention of web admin tools from questions on the site. There was no discussion of this beyond the original tag cleanup; locking - though likely well-intentioned - obscured the scope of the effort for months. Very few others were involved in any way; a handful of people did participate in retagging, but the majority of the work - indeed, the majority of editing and moderation period - over the past three months have been the work of one solitary individual. A new off-topic reason was added via comments, without review by either the community or the rest of the mod team, and was single-handedly made into the single most-used close reason on the site, in the process avoiding both the guidance given to moderators for using off-topic reasons, and the restrictions built into the system itself for creating them:
The moderator who created a reason cannot approve it himself; we want at least two people to be reviewing these before making them available. Approving a reason also activates it, and as noted above, only 3 reasons can be active at any one time on most sites - to approve more than this, an existing reason will have to be deactivated first.
I believe Hopeless had good intentions here. But by playing the maverick, he left himself and the rest of us open to criticism and without a clear defense. If he had taken just a little bit of time to talk openly about what he was doing, if he had been more receptive to criticism, this all could have been avoided.
The past couple of weeks appeared to be where things started taking a serious turn for the worst, so I went through and reopened 20-some questions closed during that period where any mention of web admin systems was clearly incidental. Some of the moderators have been doing their own reviews and reopenings as well. Going forward, we'll need to do a more structured review - at minimum, there are several hundred questions where historical locks currently prevent any attempt at community moderation; those should probably be removed before anything else. I'm open to suggestions on how to conduct a productive review of them.
In closing, I'll add that doing this analysis has been an eye-opener for me:
- The guidance we give to new moderators regarding the necessity of communication is lacking. There's an introductory email that touches on it, and of course a whole lot of history on Meta Stack Exchange... But not everyone reads the manual. This stuff needs to be baked into the UI itself, particularly when...
- Extreme outlier events should trigger something. I'm still thinking about what exactly this should involve, but for sure quadrupling the number of locked questions on a site should cause something to happen. At minimum, it shouldn't go unnoticed by...
- High-reputation users need better information. The current 10K tools were adequate back in '09, but a lot has changed since then. At minimum, these folks should be able to review the number of posts being deleted (and for what reason) over time, the number of questions being closed over time broken down by reason (including custom comments!) and the number of locked questions independant of things like migration.
- And it's probably time to revisit the notion of bespoke review tasks for those situations where you really need to get a lot of folks involved in something like a tag cleanup effort. Or... reviewing 1500 locked questions.
