Re: [Fwd: Re: [unrev-II] Meeting Summary: 22 Aug '00]

From: Eric Armstrong (eric.armstrong@eng.sun.com)
Date: Thu Aug 24 2000 - 19:35:40 PDT


Damn good post, Jack.

This paper contains an incisive and thought-provoking
analysis of the limitations of threaded discourse systems
-- but it starts by accurately evaluating their benefits.
I'm not totally enthralled with the proposed solution,
but the analysis deserves careful study by everyone
involved in the ohs-dev effort.

Comments/Highlights
-------------------
* "online environments support electronic conversations that
  expand and branch, but provide few supports for drawing
  together discourse in meaningful ways."

* they "lack support for convergent thinking processes"
  (I addressed this issue as a need for "reduction" in the
   original requirements paper. In some ways, I think that
   is a better solution -- but I've yet to identify a
   practical way to implement it! More in a moment.)

* the single-note reply option "reduces the liklihood that
  an individual will consider the option of simultaneously
  replying to one or more note"

* Even if they do, the note can only "live" in one hierarchy.
  --That's the issue that was address by the notion of a
    hierarchical document as a "view" on a network of nodes
    in the requirements paper.

* "discussions can easily go off track and individuals often
  experience a sense of 'information overload' as discourse
  structures grow and communal interest begins to fragment."
  --One reason: "A" is a statement. "B" is a reply that is
    relevant to "A". "C" is a reply to "B" that is relevant
    to "B", but pretty far afield with respect to "A". He
    calls this the "tunnel vision effect".

* "this contributes to...confusion about the intellectual
  focus of the community"

* "Hierarchical structures provide no means of visually
  displaying a sense of convergence."
  --Here, we will someday part company, if I can *ever*
    figure out how "reduction" should occur. What I have
    so far is that a simultaneous reply needs to *supercede*
    the nodes it replies to, in some significant way.
  --The issue, of course, is that not all such replies
    are summaries! So that kind of reduction is not always
    the right tack to take for a multi-node response.
  --By the same token, it should be easy for someone
    else to provide a competing summary. And there needs
    to be a mechanism for "adopting" a summary -- i.e.
    selecting, in the same way that an alternative is
    selected from an IBIS discussion (another mechanism
    yet TBD).
  --If, however, these issues (essentially interface issues)
    can be resolved, then a hierarchy can conceivably
    introduce a very *nice* mechanism for convergence, by
    *inverting* the hierarchy when the occasion demands it.

* The document focuses on "learners" in Computer Conferencing
  (CC) environments. However, researchers and designers are
  essentially learners. The processes are pretty much isomorphic.
  So all the comments he makes about learners in an educational
  setting are down=the-line applicable to an online design/
  discussion tool.

* Although he usually calls them "CC" environments, at one
  point he describes them as "asynchronous conferincing
  environments". Talk about an acronym! ACE has legs. It
  could go places. Plus spinoffs: TRACE, PACE, GRACE -- you
  name it.

* I suspect we should be basing our initial efforts around
  the analysis presented in this paper -- if not its
  proposed solution. If email is just a way to put information
  in the system, let's dispense with the concept of capuring
  email messages and go here, instead.

* "How a next-generation computer confercing system be
  designed? One appoach is to reconceptualize CC from a
  knowledge-centered, rather than conversation-based
  perspective."
  --I know Jack liked that. I find it scary, but interesting.
    Unfortunately, I wasn't able to intuit from the rest of
    the paper exactly what that meant in practice.

* "WEBCSILE represents one of several ongoing CSILE team efforts
  to support these sorts of processes across distributed
  communities".
  --sounds great. What are those efforts?
  --CSILE = "Computer-Supported Intentional Learning Environment"
    (somebody is not too hot at thinking up acronyms)

* He also mentions "an APA-style reference list". What's that?

* Each note includes a list of "Notes that refer to this note"
  --ie. backlinks. Doug will like that.

* "a 'Show Entire Thread' button displays all the notes in a
  thread in chronological order (rather than limiting you to
  a single note at a time?)
  --A really good outline display does this without thinking.
    This "one pane to select an item and another pane to
    display it" nonsense is so much GUI B.S. brought on by
    inadequate GUI tools.
   (See http://www.treelight.com/software/XmlEditor.html)
  --However, the concept of being able to see multiple
    notes at one time is absolutely, totally valid. The
    solution presented in the paper may not be the best (or
    may be the only one practical for the system he has
    devised) but it represents a very real need.

* I think the author has the view that showing the nodes
  graphically will indicate where convergence has occurred,
  and that somehow that will suffice for confergence to
  happen.
  --I'm not sure that's true, but there may be more
    utility to graphical node-representations than I think.
  --I tend to distrust that they will do the job as
    intended, without becoming overly. I'm still inclined
    to pursue the concept of "superceding" nodes in the
    hierarchy.
  --Doug gave me an example of a sitution where graphic
    networks *do* make sense -- it was an app for Bell
    Atlantic that helps people identify their clusters
    and collect them into sub-networks, or something like
    that.
  --It occurred to me then that graphic soluions
    work when there are only a few elements in the system.
    (Let's say 5-7, using that famous limitation of the
    human mind.) A person can then look at an arbitrarily
    complex collection of nodes, see patterns, and see ways
    to group and rearrange things.
  --For text nodes, though, where each node is fundamentally
    different from every other node, the utility of graphics
    is a lot less clear.
  --There is one area where good use of graphics makes sense,
    though -- prioritizing with respect to evaluations.
    Simply ordering isn't enough. A list of 5 options ordered
    from best to worst could represent 4 great ideas with
    marginal ideas and one bad one, or 4 lousy ideas and one
    good one. The system can use graphics to identify the
    rating of each node at a glance.
  --Beyond that, though, I'm not sure I see a lot of value.
    Even with categories, there are so many potential
    categories that the system easily expands beyond the
    5-7 types that will allow human pattern-recognition to
    function well.



This archive was generated by hypermail 2.0.0 : Tue Aug 21 2001 - 17:57:52 PDT