Consider one emergent discussion as a collective knitting job. Each monk can view it, appraise it, expand it. Or consider it a symphony of voices (Tolkien style?) In any case, some locus or front is implicit, with the work inexorably progressing along the time dimension, toward resolution. A piece that keeps morphing while you look at it, is probably not the easiest to work on. I see the new default order, and I'm not liking it.

Proposal: if the tree must sort by the girth of the branches, then perhaps a hybrid rep/age based view might be more manageable. Recent topics are works in progress and better viewed chronologically, IMO. Leaking the reputation is not good either, and can bias the voting. So, what if the algorithm was to pick a hysteresis threshold that is large initially but decreases along the optimal two-week voting period? A stable sort to rearrange the descendants where the reputation diff is above threshold.

But really, I think the key concept here ought to be careful pruning of information: it's not so much about the sort order but presentation. A good overview will be compact enough that the important bits do not disappear; the bits can be highlighted with the order preserved. (Also, the end of the thread is probably more visible than the middle. Do we really want to lump the noise at the end?)

Further. There was a polemic regarding some unhelpful content; opinions were put forth of suppressing said voices. Well, that would appear to be concentrating on the negative. (One wouldn't want to keep looking at potholes when riding.) But a gentler approach could be adopted, one to ensure that the good content is properly featured. Tree view is depth limited—what if this limit was adaptive? Some condition (user option) to count certain nodes as zero-depth ie never pruned if the parent is shown. A +rep node following -rep; or a +rep node following lesser rep node.
Rationale: do not cut off discussion where it probably shouldn't; good retorts or important corrections must show up.

Another thought, but this is less relevant. I have a certain distaste for trivial, back-and-forth, A-B-A-B-A-B kind of exchanges. Suppose those deep, unbranching strings of messages were flattened to one indent level (count fractionally towards depth?)

Replies are listed 'Best First'.
Re: Threaded Monks Yodeling Modeling
by jdporter (Paladin) on Jul 09, 2015 at 05:33 UTC

    I think what would help a lot is for the thread interface to be more interactive -- heck, any interactive -- such as letting the user expand/collapse branches at will, reorder the children of a node, and so on.

    I reckon we are the only monastery ever to have a dungeon stuffed with 16,000 zombies.
[reply]
[reply]
[reply]

      Expand/collapse is one option of course. I won't argue against it, but would like to point out some consequences.

      First, when you open a thread, would you like to read that as prose, or do you prefer to browse a windows registry? (Sigh.)

      Secondly, the way you read a thread will have an effect on how you contribute. A considerate monk might decide against expanding on a topic that is already large and sprawling. A collapsible tree might work differently. IOW, there is another objective that is at odds with the ease of browsing. That objective is to keep a topic in focus, topical, limited, and manageable.

      Not sure what the value of letting users reorder nodes at will would be. But perhaps, some reply nodes could be designated as subroots and appear in main sections. Almost like moving, but without breaking the tree relationship. (A placeholder hyperlink would remain.)

[reply]
Re: Threaded Monks Yodeling Modeling
by marinersk (Priest) on Jul 08, 2015 at 22:24 UTC

    I like the idea of flattening; I started becoming increasingly aware of the node consumption, as well as conversational fragmentation, when working this weekend with skosterow on File Upload On Windows 8 and Perl.

    About halfway through what is now that monstrous shrubbery of a conversation, it occurred to me that this might be precisely why some Monks make frequent reference to using their scratchpad to work through a problem. I never saw the point until this past weekend. We could have ironed out all the details, and published only the solution on-thread.

    I'll be looking into how to do that effectively for future use, but one thought I'd had is -- geez, I wish skosterow and I could have agreed to flatten a bunch of subnodes into some ancestor node and make it less painful to follow in eight months when someone else bumps into this.

    Not sure it's feasible, desirable in the spirit of the Monastery, or worth the effort given the state of Perl in the world, but your thoughts definitely mirror my own.

[reply]
Re: Threaded Monks Yodeling Modeling
by choroba (Cardinal) on Jul 09, 2015 at 21:26 UTC
    Leaking the reputation is not good either, and can bias the voting
    Why is leaking the reputation not good? I think the new default order (which you can easily change back to the old behaviour) was established because many monks thought it was good. The group of users that should profit from it is the newbies, who can't tell quality nodes from poorer answers. The group of experienced and voting monks is probably much scarcer.
    لսႽ† ᥲᥒ⚪⟊Ⴙᘓᖇ Ꮅᘓᖇ⎱ Ⴙᥲ𝇋ƙᘓᖇ
[reply]

      Why is leaking the reputation not good?
      Herd mentality? E.g. Best Nodes gives this admonition:
      By the way — Please don't upvote these nodes just because other monks thought they were good. If you do, their node reputations will increase — but then we won't really be showcasing the best nodes.

      Next, "easily change back to the old behaviour." I check on PM, and sometimes comment on basic/mundane stuff, all without logging in. No, I cannot toggle that setting.

      ... newbies, who can't tell quality nodes from poorer answers.
      One of the things I like about PM, is that you need to keep your brain in the ON position. Education is never easy, it involves some effort. I do wonder about your teaching career — did you consider your pupils as dummies who wouldn't know a difference unless told so?

      A word about the quality of nodes... My earliest PM post was this, I think. The very first reply in that thread has a rep of +20, and I will tell you that it is a rotten, rotten piece of advice. If you're looking for one CPAN module that ought to be eradicated with kerosene, then that might it!

      The pedagogically effective advice is generally highly rated though, I'll grant you that. But the rep mechanism is not up to the task. (The simplest example I can offer: say an old thread is followed up with a critical clarifying update. The first place you would check an old thread is at the end, for this very reason. But this node may get buried in the middle since nobody bothers to vote on it.)

[reply]
        I liked the idea that voting should be visible somehow. Changing the order of the replies was probably the easiest thing to implement, so we have it. It would be much nicer if the "Replies are listed 'Best First'." notice was clickable and reordered the nodes in the other way. Maybe the ordering should be separated from voting, but then I'd still like to see at least a hint (negative / low / high reputation) at each node. The question then would be how to calculate it (absolute or relative to sibling nodes etc.)...

        As a teacher, I rated every pupil's homework every week, giving them detailed explanation of why they lost 5 points here or gained 10 points there. The homework tasks, points and comments were all publicly available.

        لսႽ† ᥲᥒ⚪⟊Ⴙᘓᖇ Ꮅᘓᖇ⎱ Ⴙᥲ𝇋ƙᘓᖇ
[reply]
A reply falls below the community's threshold of quality. You may see it by logging in.
A reply falls below the community's threshold of quality. You may see it by logging in.

Back to Perl Monks Discussion