Anything untrue isn't there to be lived

Moderating LessWrong: A Different Take

[ Disclaimer: I'm pretending to be a LessWrong mod, talking to some new moderators, who don't exist. I am not involved with LessWrong at all. I'm copying what Duncan did in this post. ]

The infrastructure of your system (an online community in this case) will determine what incentives your system has. Moderators play a large role as part of the infrastructure of LessWrong. When the infrastructure is unstable or unreliable, the system will also be unstable and unreliable.

As such, it's important that our moderators are held to a higher standard than normal users.

That said, there are many different possible moderation methods. I will go over some of those here.

But first, let's cover what kinds of things we want to incentivize, particularly in the comment sections.

LessWrong is a place for shared exploration. We encourage curiosity, spit-balling, and refinement of ideas. If you imagine the commenters on an expedition together, they might propose new trailheads, point out which trails have already been explored or seem unpromising, ask clarifying questions, or try to understand each other's maps and why they're communicating them in the way they are.

If someone comments with an idea that you think is bad, underdeveloped, or unoriginal, this still counts as contributing—generating ideas at all (including 'bad' ones) is helpful at the start of an exploration. This is often why brainstorming sessions don't try to critique ideas during the actual brainstorming, as you don't want to mix generation with evaluation. It cuts people off before they arrive at their really good ideas.

As a moderator, you should err on the side of encouraging idea generation and sharing those ideas. The users will likely handle evaluating the idea, either in the form of up/down-voting and in their comment responses.

As a general rule, you, the mod, will be tasked with doing the things that feel a bit harder or unnatural. You are swimming upstream against natural incentive gradients, rather than going down them. This is what it feels like to be part of the infrastructure, rather than inside of it.

So if your natural inclination is to criticize a bad idea and downvote it, that is pretty understandable and expected.

But remember your ultimate goal is to encourage the kind of environment that fosters idea generation, idea sharing, and collaborative exploration of those ideas. And if other people see ideas getting no support whatsoever, by virtue of everyone around them thinking they are crap, the result is likely less overall idea sharing, by everyone.

That said, there is still a lot of value in pointing out trails that seem cold, unpromising, or already explored.

There is a difference between critiquing an idea out of a knee-jerk reaction that it's a bad idea—and critiquing an idea because you want to save your companions time and effort in their journey. The latter type of comment is welcome.

LessWrong is also a place for productive argument. In general, we try to encourage the kind of argument that is based in collaborative truth-seeking.

Argument is a good way to quickly identify what's important, which points are contested, and then resolve the contested points. Much like how two chess players improve by pitting themselves against one another, argument is "adversarial"—but it doesn't have to be zero-sum. The trick is trying to make it as positive-sum as possible, for everyone involved.

Your job is most difficult and most crucial in this arena, and as previously mentioned, lots of methods are available to you. It's up to you to determine what's most effective for you personally.

Some things to keep in mind:

  1. Go into every sticky situation with the attitude that something can be made better. In other words, every situation has the possibility to be salvaged, at least upon first encountering it. You may thoroughly assess it and then realize it isn't. But don't go into a thread thinking, 'ugh this is going to be a total clusterfuck,' with your hands thrown up before you even begin. You likely can do something to improve the situation.
  2. Escalate slowly. Start with gently nudging comments, clarifying questions, and keeping an open mind. If that doesn't work, try giving a warning or requesting that people take a break from the conversation. And then if that doesn't work, actually consider taking out your mod powers.
  3. Check your own triggers and biases. Whatever the subject matter, try to notice the moment it starts "feeling personal" or you start feeling a need to defend or push against. Or take a look at your recent string of comments and try to assess whether there is a pattern of potentially motivated reasoning. Ask your fellow mods for help or advice; have them look over your comments and decisions and give input. If you are trying to moderate a heated thread, it will likely only make things worse if you the mod also get caught up in a trigger or a zero-sum mindset.

In service of that, if a fellow mod points out you're driving drunk, you should consider handing over the keys. Or at the very least, check your heart rate, whether you're dissociated, or feeling angry or threatened. Maybe take a written account of the incident so you can keep track of these things, for yourself.

Ultimately, your job is to keep the space from veering into zero- or negative-sum territory.

While many conversations on LessWrong are peachy in that regard, more controversial or triggering topics will predictably lead to situations that try to turn headfirst into the red zone. You will quickly learn how to identify these topics, if you haven't already.

Here are some things you can do in those cases.

1. Be an exemplary participant

People resorting to ad hominem and strawmanning? Do the opposite. Butt your head into the conversation. Paraphrase both sides to the best of your ability. Try to figure out what their cruxes are or at least what they seem to agree on. (What's at stake for them? What is the value they're trying to protect?)

Humanize the fuck out of them.

Be curious, and ask questions, giving each side a chance to be heard.

Feel free to offer your own opinions and arguments too, but also give the people space to feel heard, especially if they're triggered. Don't try to switch back and forth between being curious and then offering opinions. Do one thing at a time. Choose your moments to share yourself prudently. Less is more, here.

2. Go meta, point out what's happening, bringing everyone's awareness one level higher

Works best if you try to only say true things that both sides would agree with.

"It looks like userA has mostly been trying to get across the point that ____, from different angles. And it looks like userB mostly responds to this by reiterating ____. My guess is that you two are talking past each other in this way: ____."

From there, you can propose some new tactic or course of action.

You might even go further in your own evaluation of the situation, if you got models to back it up.

"I see that this argument has been going on for about 5 hours, and not much has progressed in that time. My personal sense is that another hour or two of this will not produce significant, forward-moving change."

And then you can propose they take a break, have a facilitated conversation over Skype, or evaluate what they're really wanting out of the conversation and see if there's an easier way to get it.

3. Verbal up-votes and down-votes

As a mod, you probably already have greater up-/down-vote power than most, but your up-/down-votes are also anonymous. And sometimes you want a little more weight behind your actions. After all, your actions are the walls and foundation that this whole thing is getting built on.

There are many reasons you might find yourself not wanting to get super-involved in a thread (by doing 1 or 2, above). Maybe other mods are already handling it. Maybe you don't have time. Maybe you know the topic will mind-kill you. Maybe this is part of your overall mod style.

So rather than getting embroiled in the conversation, you can express your opinion in smaller, nudge-y ways, that don't further engage you. But do make it clear that you have an opinion.

I think some of the best times to use this, in particular, is when a user has clearly crossed a line.

For instance, if a user resorts to ad hominem ("you're an idiot"; "i can't believe they let you post here"), it is probably clear to most people reading it that it is an ad hominem. But just leaving it to rest in that assumption can still create ambiguity around "what we do about that around here."

This is the kind of situation where social signaling is quite useful.

You don't want to casually, socially downvote a thing that is ambiguously 'bad'. (Like when someone makes a point you disagree with.) In those cases, you want to be articulate and just say what your disagreement is. Otherwise, you come off as trying to use social status as an argument tactic. We do not encourage this.

But in cases where the thing is pretty obviously a line-crossing, it is helpful to "step forward" and casually dismiss it or openly "swipe left." This demonstrates to the social sphere that lines that are crossed will not be merely passed over and ignored. The main benefit is to the other readers and community fabric, and less so directly to the person you're responding to. But it's a key step in keeping your garden free of weeds.

I recommend socially savvy ways of dismissing ad hominem. Less being scolded by your principal (this comes across as "too much"); more like taking a scalpel and taking them down one notch.

Only one person needs to respond this way to each instance of line-crossing, and it doesn't have to be a mod. (It helps if the person is well-respected.) One is enough. Any more and it can feel like a mob.

Sometimes, mods/users might try to give them the benefit of the doubt by trying to engage them, ignoring the ad hominem as though it didn't happen. My guess is that there is a strong pull to "salvage the discussion" and this is a tactic to try to do that.

As a mod, you need to have your eye on what kind of incentives you are creating / dis-creating. Ignoring line-crossings is going to be counterproductive in the long run because, each time you do, you demonstrate to everyone else "this is how we do things around here." This encourages more line-crossings.

Sometimes, I see mods/users explain the offense in detail and why it needs to be discouraged. If the offense is truly an ambiguous case, this is a good idea. But if it's one of those things that should be obvious, try not to over-explain. Over-explaining signals that the offense REQUIRES explanation.

Sometimes this is the case. But when the line-crossing should be obvious ("you're being a dickweed"), do not spend MORE effort than the original commenter in dismissing it. Not only do you want to signal that this is a line-crossing, but you also need to signal that it's an obvious line-crossing.

This lets people not waste cycles on every troll, every asshole. It is OKAY to ignore and dismiss people who are going to suck the positive energy and momentum out of a convo. Your job is to maintain that positive-sum-ness. Openly drawing a boundary can save everyone time and energy.

If you're actually not sure if it's obvious, it's probably not obvious. If another mod disagrees with its 'obviousness', it's probably not obvious enough. If it's not obvious, don't treat it like it is. This is dishonest.

Does this carry a risk of ostracizing some people unfairly? Yes. A trade-off will be made, one way or another. The question is where you want the trade-off to be. I recommend deliberately choosing it.

Let's go over a less obvious case.

In the comments of Duncan's In Defense of Punch Bug article, SilentCal made a comment wondering what was asymmetrically unfair about the game of punch bug.

I don’t see what’s asymmetric about the ‘no punch back’ rule at all — the punchee is free to spot the next bug, in which case they will become the beneficiary of the ‘no punch back’ rule.

Another user named benquo (who is prolific, well-respected, and reasonable in general) responded with:

Is it hard for you to imagine that some people might not be violent sadists?

When reading that comment, I could feel myself "taken aback" by the force of the phrase "violent sadists."

There's a sideways implication that Duncan or others who enjoy punch bug are violent sadists, but this is not a clear, unquestionable instance of someone being personally attacked, and it is possible that the implication wasn't intended.

I would classify this as a strong framing with a leading question. I get the sense benquo had a "tighter grip" on the frame than might be ideal for collaborative truth-seeking. (He could be holding a valuable truth, but the accusatory tone doesn't invite others to engage freely and openly.)

If benquo didn't have a large body of engagement that demonstrated clear and insightful thought and instead had a history of being consistently problematic, this would be a slightly simpler case.

As it is, I personally think tactic #3 (verbal down-vote) won't be as good here, and it's better to try #1 (participate) and/or #2 (go meta).

  • Benefit of the doubt is warranted because of past evidence. (Note: if this were a relatively new user, I'd also give benefit of the doubt.)
  • When someone presents a strong frame (especially a moralistic one), this is often a sign of something important being threatened. They might feel they, personally, are being threatened. Or they might sense large amounts of value will be destroyed in some way.
  • It basically never works to try to convince someone with a strong frame that they are NOT under threat or that value will NOT be destroyed. ("Punch bug is pretty harmless tho.")
  • Trying to back this person into a corner also tends to be a bad idea. ("You realize you're being unreasonable, right?")

What to do?

Well, look, as a mod, you need to keep your own head above the water, first and foremost. Put your own oxygen mask on, etc. Maybe take a few deep breaths. Good? Proceed.

The more things start to escalate, the more you're going to want to use #2, versus #1. Going meta tends to break the ice a lot better and stop whatever doom train you're all on. That's your judgment call.

If I felt like trying #1 though, I might try to get a better sense for what this person feels is under threat. Once I personally manage to "be in that headspace" (which is where you'd try to be to pass their ITT), I can then humanize them. ("Oh, huh, if I put on that frame, I also feel _.") This move is a little like steel-manning, but you are trying to find the place IN YOURSELF where you can feel the way they feel and see the way they see. And then you share the results of that internal exploration.

One way to create positive-sum-ness here is to try, as much as possible, to reduce adversarial vibes (and feelings of us-vs-them or feelings of desperation).

So, create more breathing room in the conversation. If someone feels threatened, reassure them that they at least aren't alone in their feelings. I think it's important to share the emotional impact here (even indirectly), rather than just straightforwardly agreeing with someone's viewpoint.

Avoid making them feel targeted, mobbed, attacked, or cornered. If you notice people trying to pile on, well... OK, LessWrong unfortunately doesn't have great mechanisms for slowing that down. You can try to go meta (#2) and point out that you think this is happening and ask people to slow their roll. This might work. If not, you might have to 'suspend' the entire thread, if it gets out of hand.

A decent way to create breathing space is to actually take the convo off LessWrong and into a private space, such as email. You will have more control here and less risk of escalation. You can PM the user and ask them to pause the LW convo for a while and first hash things out one-on-one. I recommend having this as a regular tool in your toolbox.

(Ultimately, it sounds like a mod did engage benquo one-on-one, and I think that's probably the correct call in this instance.)

More on creating positive-sum-ness

This is distinct from "trying to seek truth" or "get to an accurate conclusion."

As a moderator, you are the walls. You are the infrastructure. If you just engaged in trying to get to the truth with everyone else, there would be no walls, and you all fall into chaos.

What this means in practice is that if someone is acting out in the comments, your job is not to try to argue with them and seek "what's true." Your job is to create the incentives that lead everyone to be better aligned with truth-seeking.

You can do this by example. But this is a means to achieving the greater good. It may not always be the best means, and you need to be asking yourself that.

If there is a particular truth you are seeking, it is the answer to "what is the best way to create and foster a community of truth-seekers"? In every action and movement, you are trying to answer this question, for yourself and for the other mods and for the community as a whole.

Yes, this is all terribly meta. But if you get stuck in the object-level questions like everyone else, you aren't playing the mod role anymore. You're just another user.

Be the wall we need; not the wall we deserve.