Why ThinkLab uses a flat discussion structure (not threaded)

In another thread @jackpark suggested there may be benefits to bringing more structure to the ThinkLab commenting system. I believe one of his points raised the idea of structuring ThinkLab comments more like Slashdot or Reddit:

1) make use of natural outline structures to as many depth levels as required

There's been a debate about flat vs threaded commenting systems for a long time. For our use case I'm pretty firmly in the camp that says flat is better. Here's why:

• Whenever someone posts a comment we want it to be a well thought out comment in consideration of all the comments that have already been made in the discussion. A flat structure supports this.
• We want to keep discussions focused on one subject. This increases the signal-to-noise for anyone who is looking at the discussion because of an interest in the subject. A threaded discussion structure tends to encourage conversation to go all over the place. If people have an important point to make on a new topic we'd like them to post a new discussion with a new subject.
• When people are following a discussion we want to be able to quickly point them to the new comments they haven't seen yet. This just gets really complicated in a threaded view.

I think that I did a rather poor job of explaining my thesis; I was certainly not advocating mimicking slashdot or reddit. Rather, I had in mind something closer to what Jeff Conklin has to say about the matter: https://www.youtube.com/watch?v=pxS5wUljfjE

Clearly, keeping dialogue focused on one subject is a useful goal. Let me build on that for a moment: A particular thread is about a single topic; this thread is about the difference between flat and structured dialogue. It necessarily entails, at the very least, two distinct subjects: flat dialogue, and structured dialogue.

But those topics, as indicated in the bullets, entail signal-to-noise ratio, and findability.

Point of interest: a claim was made in this "flat" thread; it stands out as if it needs its own specific let's call it "sub-thread": "A threaded discussion structure tends to encourage conversation to go all over the place."

When one makes such a claim, it is considered good form to offer citations, or at least some level of evidence; a conversation about the conversation platform is no-less important than conversations about, say, hemochromatosis and leukemia.

My inclination, were this to be a threaded conversation, would be to launch a "con" node, that is, an argument node which, by its node type, signals disagreement with that claim. I would do so from the observation that a single thread always entails other subjects, and that a proper structure keeps those subjects bound to their context. To do that in a flat thread, as here, requires a lot of verbiage to keep things flowing, lowering, I would argue, the SNR, forcing others to read more prose than could otherwise be signaled by structure and typed nodes.

The specific vocabulary of such a structure is one created by Jeff Conklin for what is known as IBIS conversations. Neither Slashdot nor Reddit use anything like that.

I watched the linked video and I get it. I can see that regardless of how we display the conversation (flat or threaded), behind the scenes there could always be some structure extracted and mapped. I also get the discussions could be broken down into node types of ideas, questions, supporting arguments, disagreements, etc. But the question is so what? The question is not whether a structure exists, but what user interface best works to accomplish our goals?

My inclination, were this to be a threaded conversation, would be to launch a "con" node, that is, an argument node which, by its node type, signals disagreement with that claim.

Right, I get the concept but what do you mean by "threaded conversation"? Reddit and Slashdot have threaded conversations.

In ThinkLab's current flat structure you could signal disagreement like so:

A threaded discussion structure tends to encourage conversation to go all over the place.

I disagree, ...

I'm not saying there won't be some benefits to structuring things more. I would just ask you to try to think about what specific user interface changes you would suggest. And it would really help if you can point me to some other popular and easy to use sites that have implemented them. I've taken a look at DebateGraph and I must say this interface seems far from user-friendly.

ThinkLab already has a huge challenge ahead of us. We're basically trying to reinvent science. Reinventing how comments work on the internet is really asking too much. But if you have any specific ideas that would improve the site right now let me know! We are currently in the process of adding a discussion labelling system which will work very similarly to what has been implemented for GitHub issues. There will be a separate post about that.