Taking it Private: Short Circuiting Demon Threads (working example)

by Raemon4 min read22nd Jan 201870 comments


DisagreementCommunication CulturesDemon Threads

This post is intended as a working example of how I think Demon Threads should be resolved. The gist of my suggestion is:

Step 1. Make it easy and common to take a conversation private if someone is feeling annoyed/threatened/angry/etc (if it seems like the conversation is actually important. Meanwhile, also make it easier to tap out if the conversation doesn't seem like the best use of your time)
Step 2. In private chat, two people do their best to communicate honestly, to notice when they are defensive, to productively find the truth as best they can. (I think this is much easier 1-on-1 than in public)
Step 3. Someone writes a short summary of whatever progress they were able to make (and any major outstanding disagreements that remain), focusing primarily on what they learned and rather than "who's right."
The summary should be something both parties endorse. Ideally they'd both sign off on it. If that trivial inconvenience would prevent you from actually writing the post, and you both generally trust each other, I think it's fine to make a good-faith effort to summarize and then correct each other if they missed some points.
Writing such a summary needs to get you as much kudos / feel-good as winning an argument does.
Step 4. The public conversation continues, with the benefit of whatever progress they made in private.
Ideally, this means the public conversation gets to progress, without being as emotionally fraught, and every time something comes up that does feel fraught, you recurse to steps 1-3 again.

Qiaochu had a criticism of the Demon Thread article. I had said:

Demon Threads are explosive, frustrating, many-tentacled conversations that feel important but aren’t.

He responded:

I want to object to this framing, particularly the "but aren't." It's far from clear to me that demon threads are unimportant. It may seem like nothing much happened afterwards, but that could be due to everyone in the thread successfully canceling out everyone else's damage. If that's true it means that no one side can unilaterally back down in a demon thread without the thing they're protecting potentially getting damaged, even while the actual observed outcome of demon threads is that nobody apparently benefited.

I initially responded publicly. (I think the details are important in their own right, <linked here>, but aren't the main point of this post)

We still disagreed, and the nature of the disagreement hinged on past threads full of social drama. This was exactly the sort of thing I didn't want to discuss publicly on the internet. Yes, the details mattered, but public discussion would have lots of bystanders showing up with opinions about the object-level-details about the social drama itself.

In this case, Qiaochu and I were able to discuss it privately, which:

  • helped my own demon thread model
  • was a useful working example of what should come out of steps 1-4.

So I've written this up as a post instead of a comment. I haven't run this by Qiaochu yet (I think getting formal permission/endorsement adds an "significant trivial inconvenience" that might disrupt the process too much), but I expect him to endorse the following, and I'll update/clarify if I got anything wrong.

Things I learned

i. Mattering-ness is orthogonal to Demon-Thread-ness

The most important update on my part. Qiaochu provided a few examples where it felt right to call a thing a demon thread, which the thing-in-question mattered in some sense - either because the tribal affiliation and status mattered, or because the actual ideas getting discussed mattered.

"Is it a demon thread?" is more about "there is some weird force compelling more and more people to argue, raising tensions" than it's about "the argument is counterproductive or going in circles" (although I think the latter is common).

Since matter-ness isn't part of the central definition, I've removed it from the description at the beginning of the post.

ii. Avoid bundling normative claims with descriptive claims.

One reason I think Demon Threads (often) don't matter is a normative claim about what people should value, and it is unfair to bundle this with claims about what people should do given what they currently value, even if I think they're being silly.

Conflating descriptive and normative claims can be a useful (but deceptive) rhetorical trick, and part of the point of LessWrong is to avoid doing that so we can think clearly about things.

Empirically, people care about what groups and ideas have relative status among their peers.

My point was more like: Arguing on the internet about the relative status of things is not effective altruism. People care about things other than accomplishing the greatest good for the least effort. It is perhaps most of what most people care about. And that's fine.

I think this claim is still relevant, because people often seem to think the thing they're doing is helping a much larger amount than it is, as well as accomplishing different things than they think it is. (i.e. you think talking about the president is having an impact on national policy, but it's mostly having an impact on what opinions are acceptable to express in your local peer group).

I do think, in most social/political-drama-laden threads, if people took a step back and thought about it, they would either realize an internet debate wasn't the best way to accomplish their goals, or they'd realize their goals were different than they thought they were.

iii. Maybe something didn't matter before the demon thread, but after a giant explosion of arguments happens, it may matter a lot (at least to the people involved).

I cited an example where, in a local community, people started arguing about [internet drama from several years ago]. Prior to the argument, it hadn't mattered what your opinions about that particular political drama was. But suddenly, everyone knew what many prominent community-member's opinions were, and people disagreed strongly, and there was a risk that if the thread went the wrong way, having one set of opinions might no longer be okay.

Qiaochu and I agreed it would have been better if the argument never happened, and that the political drama wasn't objectively important. But he argued, once it had exploded, it became relevant to the people involved. So the pressure to add your 2 cents was real and important.

This seems true, but also feeds back into my central claim, which is that it's best to stop malignant demon threads before they begin.

Outstanding Disagreements

Often, when people are coming from very different intuitions, they can argue a lot about factual claims, and agree that each other made good points... and still go back to basically holding their original position.

This can be frustrating, but understandable: people have a lot of background experience that feeds into whether something makes sense. Explicit arguments often can't fully address that background experience.

While we agreed with many of each other's claims in principle, each claim of Qiaochu and I included lots of words like "usually" and "sometimes", that were doing a lot of work, and our respective takeaways rounded those words in directions closer to our original positions.

Qiaochu's current overall position as I understand it is:

People are constantly tracking the relative status of groups and ideas, our intuitions about this are actually pretty good - both at detecting what's going on, and whether it is relevant to our goals.

(I originally interpret this to basically be arguing against the "We're adapted for Dunbar Number tribes, therefore our intuitions for the modern world are useless" hypothesis, which seemed confusing. In the comments below Qiaochu clarified among other things that although our society is bigger, so are our tools for broadcasting signals. See his comment for more clarity)

My current position is:

People's intuitions for tracking the relative status of groups or ideas is doing something, but it's not really doing what they think it is, and it's not well adapted for the modern world. Lots of things matter as much, or more, than political goals, but we have a much easier time understanding (or thinking we understand) political goals, so we spend disproportionate time on them.
Meanwhile, because the modern world is different, accomplishing political goals that are relevant outside of your immediate social circle usually requires doing things that feel counterintuitive.