Ka-Ping Yee, Marti Hearst, (2005) A Visualization to Facilitate Productive Discussions

---

Comments by KushalDave

This is a problem near and dear to my heart, and you do a great job setting it up and describing related work. The interface itself is gorgeous and intriguing! It uses space efficiently and has a lot of pontential to represent discussion information concisely. I especially like the subtle touches, like bolding the first line of each paragraph and calling out URLS. It seems like it will definitely make skimming easier.

The issue of quick navigation is certainly a tricky one. I wonder if you could do something where mousing over a particular section of the discussion would expand it and shrink its parents and siblings to rectangles, freeing up space for reading
without heavyweight diving? Or, maybe you could support linear scrolling and have the screen zoom in and zoom out appropriately?

I guess one of my biggest concerns is that this still seems tied to the the idea of well-behaved threads. If you're tying this to mailing lists, it seems like you're especially likely to get what I call "brain dump" posts, where users discuss several contentious subpoints and maybe introduce some new ones, all in a single message. Not only would it be misleading to thread them in any one place and mark them as purely positive or negative, but long messages don't seem well-suited to your interface. Although the +/- convention adds power, is this actually that widely used in practice? It might be more interesting to explore a purely web-based forum that made voting much simpler and integral to the process. I'd love to see more ways of moving between threads - what if you used some of the empty space to show messages that appear related by the content of the message? Maybe users could have more power over the structure of the artifact?

I love the idea of using something like width to reflect some other dimension. Maybe it could be a combination of the known authority of the author (like karma in Slashdot, or seniority in python development) and the voting on that particular message?

KushalDave

---

Comments by LiangZhou

The layout technique (reminiscent of a treemap) described in this paper brings out an important property of threaded messages, where later messages respond to much earlier ones, not following the timeline. It property makes automated discovery of question-answer pairs more difficult than in traditional conversations (like meeting recordings) where no large gap in time and no interweaving messages between speakers occurs.

LiangZhou

---

Comments by AdelePontes

Our work shares some common goals and issues. You claim that "while there is much value to be gained from a structured representation, we feel that too much emphasis on the classification of arguments and their relationships impedes flexibility and usability in practice". Likewise, we have evidence (from a decision-supporting prototype tool that follows the IBIS model) that too much structure may be declined by users (or even inadvertently misused). We also share the goal to support users in interpreting and keeping track of the discussion, with an emphasis on supporting their active participation.

We also believe that users should be able to "use their current tools and abide by current practices". While you have used e-mail tools, we have used different forms of discussion forums. Could you characterize the range of purposes supported by e-mail as the tool for discussions? For instance, if a major goal of the discussion group is to make sure that the discussion is archived (for purposes of maintaining an organizational memory or granting wide access to the discussion), have you investigated to what extent e-mail tools would be adequate?

While you have added +/- markers to the conversation exchanges, we have been exploring the use of markers as well, but using a different set. Because our main goal is to support scientific debate, in which polarized discussions may not be the norm, we have decided to include varied rhetorical markers such as ask/answer, inform, explain, comment, exemplify, elaborate, and so on, while still including agree; agree, but...; disagree, but...; disagree as in decision-making exchanges. As in your case, users are not required to qualify their communicative exchanges using these markers, but we believe they would benefit from doing so when reading/interpreting the resulting discourse.

Have you investigated what kinds of discussion or genre benefit most from the +/- approach, and what kinds might benefit from more fine-grained or sophisticated markers? For instance, in a scientific debate in which achieving consensus is not necessarily a desired goal, do you consider that there might be a different set of markers?

AdelePontes

---

Comments by MartinWattenberg

I like the elegant interface that is presented in this paper. I'm impressed by the level of care that's gone into the design--for instance it's terrific to see a display that's specifically meant to make it obvious when people aren't backing up their arguments with evidence.

It's interesting to think how the interface would scale to larger sizes. Would it feel cumbersome to have to keep clicking to get to deeper parts of the tree? Kushal's suggestion of a focus+context navigation seems good to me. Paula Newman (see http://www.visi.com/~snowfall/DA_Newman.pdf∞ ) discussed a related style of layout called a TreeTable and described a fisheye view approach.

One problem I've always had with online discussions is that they are static--once a message is posted, it's stuck there forever. I'd be very interested in some sort of wikified discussion that allowed editing of earlier posts. With the right sort of social norms I think this might produce very polished, high-quality debates. I'd love to discuss this further at the workshop!

MartinWattenberg

---

Comments by JudithDonath

I like the idea of having some ways of very simply marking up a discussion. These are lightweight enough to not interfere with the writing, but add useful information. I think it may be useful to take this further. For instance, if you assume that people will label their responses as supporting or against with an initial + or -, why then attempt to deduce what is a “question” from the punctuation on the first sentence. It could be a rhetorical question; alternatively, a posting requesting clarification may start with something other than a question. Why not just ask that people who want their posting to appear as a question start it with an initial “?”.

Indeed, once you do this, the system design become that of saying what are the special layouts we want to provide to people for the sake of structuring discussion and then giving people a lightweight way to mark up text entries to denote them. It then also makes sense to think about why you want to use markup rather than an integrated editor. I’m assuming that it is because you want this to be useable for all platforms, including someone just using their regular mailer. In that case, you also want to design a version where they can see at least some of the structuring, probably in the way the subject lines and/or top of the text are displayed.

---

Comments by BrendanOConnor

I'm impressed with the well-thought-out design in this paper. The clean visualizations of argument paragraphs already look useful to me. However, I am wondering how the input system works: do people need to classify every paragraph they write with the [+] and [-] symbols? It would be interesting to see an input/authoring interface (as suggested when you say you may be adding a web-based posting inteface) in the same design as the display screens shown in the paper here. Could there be a radio selection of [+|-] ? Or other simple annotations?

It would be interesting to see how search or simple text analysis could be incorporated. For example, one could a thread to a much older one about the same topic, though the authors may be unaware that old thread existed.

-- BrendanOConnor

---