The blog post "A theory of moderation" was written by Jeff Atwood in 2009. It is prominently linked in every election, and often referred to as kind of an official document on the role of a moderator on SE sites.

After being a moderator for a long time, some things bother me every time I read certain parts of that post. The main one is also quoted most often in my experience:

Moderators are human exception handlers, there to deal with those (hopefully rare) exceptional conditions that should not normally happen

and the related introductory sentence:

But what do community moderators do? The short answer is, as little as possible!

This sounds nice. The problem with that statement is that it is simply not true. The moderators on Stack Overflow handle thousands of flags a day. There is nothing exceptional about a large amount of the flags any site receives. The blog post even confirms that by stating "A lot of the moderation work is extremely mundane, almost janitorial.", which seems to contradict the previous statements in that blog post.

Moderators act in many situations that are in no way exceptional, most of the flags are routine. The "exception handler" analogy has some merit, the most important job of the moderators is to handle anything that the community can't handle easily itself.

But the implication that this is rare is misleading, there are a lot of common issues that simply can't be handled by the community. And on smaller sites even the parts that the community could handle theoretically, often need moderator intervention as there isn't a critical mass of users with the right privileges yet.

Answer deletion is almost entirely handled by moderators on most smaller sites, community deletion is uncommon there. And even on sites where the community regularly manages to delete non-answers, there is no harm in moderators removing such posts when they encounter them.

The ideal moderator doesn't do "as little as possible", they help to clean up their site to the best of their ability. They certainly should know when to refrain from acting, but I do think that this description on the blog post is misleading.

Some other details that probably should be changed is the part about contacting users via email, and maybe some updated screenshots that represent the current flagging dialog.

In summary, I think the blog post is contradictory in parts and should be clarified. There are a lot of good points in there about moderation, the specific advice is generally good. But it is put into a narrative that I find misleading, it gives the impressions that the moderator job is a lot more "hands off" than it actually is.

  • 2
    I think that the "as little as possible" is that the system is designed in such way that most of the tasks can/should be dealt by the community and moderator intervention is optional at best.
    – Braiam
    Commented Dec 21, 2014 at 14:22
  • 5
    @Braiam That's the theory. In practice, on a large site LQ Review Queue can be a big flag-generating machine, producing moderator-only flags for disputed reviews when the reviewers are conflicted or the answer has positive score.
    – user259867
    Commented Dec 21, 2014 at 20:40
  • 8
    Note: It would probably be a bad idea to change the original post; rather, post an updated, lightly rewritten version and link to that in the original and wherever else in current FAQ/help/elections/etc the post is being referenced. Having a historical record of "this is what we used to rely on" is much better than rewriting history. Commented Feb 22, 2016 at 21:11

1 Answer 1


The critical take-away from that post is summarized in the closing:

The ideal moderator does as little as possible. But those little actions may be powerful and highly concentrated. Judiciously limiting your use of moderator powers to selectively prune and guide the community -- now that's the true art of moderation.

This is the essential theory of moderation: when it's done well, it's hard for most users to tell that anything is being done at all.

This implies a few things that, unfortunately, are not always true:

  • The majority of posts will not need to be moderated
  • The bulk of moderation won't be done by moderators
  • The majority of a moderator's actions will not be controversial (in the sense of dividing the community) - rather, they will be exemplary, implementing the community's wishes in a way that is understandable.

It's not hard to find a site where one or more of these ideals isn't met. But those remain the ideals that we should be working toward. From that same post:

We designed the Stack Exchange network engine to be mostly self-regulating, in that we amortize the overall moderation cost of the system across thousands of teeny-tiny slices of effort contributed by regular, everyday users.

On a site that gets a question or two a day, this is mostly just talk; a nice ideal perhaps, but not an essential one. But on sites that get hundreds or thousands of questions each day, with answers to go with them, this ideal is what allows that to keep happening! I wrote about this at length a couple of years ago:

By rights, Stack Overflow should have died already, turned into an irredeemable cesspool by a combination of outsider influx and insider burnout. You can argue (and many do) that we're headed that way - but we've been headed that way since day one. The best we can hope for is a stable orbit, forever falling but never crashing. I believe there are two major reasons why Stack Overflow has managed to scale far beyond the expected limits of a group:

  1. Conversations not required. When a question is asked on a traditional forum, answering it often demands some amount of participation from at least a portion of the community. Details are fleshed out, the problem is clarified, solutions are proposed and debated, others with similar problems chime in with their experiences, tangential points are made, and eventually - anywhere from hours to months later - the conversation dies out. It's a very social, very natural way to interact. And it suffers mightily from the problem that Shirky talked about: all that back-and-forth and associated latency kills any hope of scale. On Stack Overflow, we close or delete questions that can't be answered straight away - it's not very sociable, but it scales wonderfully by effectively enabling a vast, human-powered computational grid.

  2. Tools that allow decoupling moderation from communication without separating moderators and users. While Stack Overflow does have a powerful "moderator class" elected by the community, a fairly large portion of the actual moderation is performed by individual members of the site, those who've participated enough to demonstrate sufficient familiarity with the community. While this has been a fundamental part of the system for a very long time, I didn't fully appreciate how it relates to scale until I started working with very small Stack Exchange sites: the proportional cost of moderation is much higher, even though the total volume of work is lower. Many hands make (relatively) light work... As long as the system puts tools in those hands.

Moderators on Stack Overflow handle thousands of flags every day, but those represent only a fraction of the flags raised (or about 15% over the last month). Thousands of posts get flagged every day on Stack Overflow, but only a fraction of the posts created get flagged. And while you can usually find a discussion or two about some moderator action on Meta Stack Overflow, the vast, vast majority of moderator actions go unnoticed and undisputed.

Routine exceptions

The problem that folks usually run into with that blog post is twofold:

  1. They don't read the whole thing. No, really, they don't. That's why I started here by quoting the ending, because I'm pretty sure an awful lot of the folks who quote from it don't make it that far.
  2. They don't realize that exceptions happen all the time. If 3 out of every 100 posts might need a moderator's attention, that doesn't sound like a lot... Until you're getting 600K posts every month.

"Exceptional" and "as little as possible" don't mean that moderators shouldn't act when necessary; rather they mean that we've hopefully designed the system such that most of the time it won't be necessary for moderators to act. That design is an ongoing process, and sometimes we fall well short of that ideal... But that's the goal we should all be working toward: the folks designing the system, the folks using their moderation privileges to help shoulder the load, and the moderators themselves. Because when you see one person doing too much, that's a sign that someone else isn't doing enough: the balance is upset, and the only solution is for more of us to step up and help.

Something to work toward

In conclusion, no I don't think we need to update that post, or the help center page that mostly mirrors it. It's the theory on which this whole system is built, and more importantly it's a set of practical goals to keep in mind when we continue building - whether that involves new tooling, new rules, or new guidelines. It may not be particularly useful as a low-level instructional guide, but those are better left elsewhere; this is "A Theory of Moderation", not "A guide to moderating comments" or "why is everything on the Progse homepage closed?" after all.

And as mission statements go, I'd have to say that it's remarkably grounded.

  • 1
    In short, you're emphasizing the word "Theory" in the title of the post. It describes a theory from which goals are derived and used to guide decisions about how things should work, rather than being a guide to how they necessarily do work in practice.
    – WBT
    Commented Jun 21, 2019 at 17:48
  • Well, yeah, @WBT - accepting the status quo is implicitly a race to the bottom; if you don't set goals that aspire to something better, build and test theories that aim for an improvement... Your community is destined to rot.
    – Shog9
    Commented Jun 21, 2019 at 17:52
  • I think it's good to have an aspirational piece which sets goals, and I even agree with the specific goals laid out in that piece. I think OP's concern is that the blog post isn't framed that way. For example, it says "What do community moderators do?" and "If you are a community moderator on a Stack Exchange site, here’s what to expect" using present-tense indicative language without qualifiers like "ideally" (or "should" instead of the first "do.") It also uses the phrase "rare exceptional conditions" when, as you've pointed out, exceptions are routine. The framing can be a bit misleading.
    – WBT
    Commented Jun 21, 2019 at 18:04

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .