Thinklab Meta [meta]

Should we aim to keep discussions narrowly focused?

I'd like to know if people agree with my assessments below:


It seems to me there are several good reasons to keep discussions narrowly focused.

  1. Better signal-to-noise ratio. People will be attracted to reading a discussion based on the subject written for it. If they encounter conversation that doesn't match the subject this will likely lower the perceived signal-to-noise ratio of the thread.

  2. Minimizes confusion in thread. The introduction of multiple ideas within a thread is more likely to lead to confusion as a result of parallel conversations happening at the same time. This is one of the downsides of a flat (vs threaded) discussion structure. But there are reasons ThinkLab uses a flat discussion structure.

  3. Prevents ideas getting lost. Potentially high value ideas can get "lost" if they are posted in a discussion with a subject that doesn't capture the idea. (Potentially interested users won't see the idea when it's posted and future users won't find the idea when they search for related keywords.)

Implications for user behavior

  1. If someone has multiple points of feedback that could each lead to substantial discussion then they should post each point of feedback individually so that we can have a separate discussion thread for each. For example, it may have been better for @alizee's post here to have been separated into two.

  2. If someone has multiple minor points of feedback that are unlikely to lead to much discussion they should go ahead and post them together. An example would be @dhimmel's post here perhaps.

  3. If someone is introducing a new idea to an existing thread that is likely to lead to substantial parallel discussion they should post this idea in a new thread (and perhaps link to to the new thread from the thread they were going to post in). An example is @jonathanwren's comment here.

Possible ThinkLab changes

  1. When people click to post a new discussion we can offer some advice regarding keeping discussions narrowly focused.

  2. We can offer a "reply as new discussion" feature which will create a new thread and automatically add a link to it from the current one.

What does everyone think? @dhimmel have you had any such thoughts in relation to the discussions in your project?

Also, just FYI — I don't want to stress anyone out about how to use the site. This is all an experiment and everyone is figuring it out as we go!

I see both sides of the issue. Our most valuable discussion thus far includes 21 posts from 5 users. The initial topic was general and recruited important suggestions, however specific conversations arose where separate discussions would have provided a clearer picture. However, I'm not sure these divergent topics could have been anticipated beforehand.

Specificity trade-off

I see a specificity trade-off that:

  1. If discussions are general, the posts within need organization — I have been using headings and user mentions as a partial solution.
  2. If discussions are specific, the discussion list can become unwieldy. A search/filter feature as well as tags in the discussion menu could help.

I have found extensively hyperlinking topics in my posts helps to alleviate both these issues (example).

Post dependency proposal

If discussions stay general, I can imagine a dependency linking feature between posts could be helpful. For example, an option would enable me to add your initial post as a dependency to this present post. In other words, to understand my post, one should first read it's dependencies. When viewing a post, a user could ask to collapse all previous posts in the discussion that are not part of the dependency chain. The benefits of flat discussion would be maintained, while incorporating advantages of a threaded model.

Reply as new discussion

The reply as new discussion feature is a great idea. We have already manually taken advantage of an analogous feature by hyperlinking. See @jspauld's post here. Therefore unless you see a great user experience benefit to the dedicated feature, I think it's a community standards and education issue mostly.

There might be pitfalls that we are overlooking, but I think the extra discussion structure is a great idea. We might fulfill the needed behavior with a simple structural design:


  • Any comment can have many children discussions.
  • Any discussion can have one parent post.


  • Simple.
  • Encourages flatter discussion by letting people track the proper background needed to understand what is going on for a given discussion.
  • Lets an OP structure off the bat a subject (s)he wants to tackle, by having one "general" discussion whose first post spawns out several "sub"-discussions.


It doesn't implement the comment dependency feature that @dhimmel is mentioning above, which I think is a good idea in general but an overkill for now. Especially, as a result of this proposal, the discussions should be pretty self-contained and dependencies of a post should be in the discussion itself, above the post of interest - or else mentioned in the body of the post.


  • Engineering: the simple dependency structure [discussion -hasOne-> parent post || post -hasMany-> children discussion] should be pretty straightforward to implement.
  • Design: having a small UI per post, linking to the children discussions, and another UI per discussion, linking back to the parent post should be sufficient. Creation of the post/discussion links could be done in the edit view of a post, after the discussion is created, or in the edit view of the first post of a discussion (back to the parent comment). The bigger challenge is how to display this inter-discussion dependency when browsing discussions. We might want to hold off on thisvisualization for now.
  Discussion System
Referenced by
Cite this as
Jesse Spaulding, Daniel Himmelstein, Antoine Lizee (2015) Should we aim to keep discussions narrowly focused?. Thinklab. doi:10.15363/thinklab.d64

Creative Commons License