REALM: Kushal Dave, (2005) Restructuring online conversations

paper∞

The web version∞ is back up!

You can get an Excel file of the results here∞

---

Comments by LiangZhou

This paper describes an interesting interface for reorganizing online discussions, especially for online debates or decision-making related conversations. I found it intriguing using previous responses to trigger more future discussions.

There has been recent development from the Natural Language community on summarizing online discussions. There's always the assumption that whenever there is debate, there are two groups of participants, one supporting the claim and the other against it. It has been difficult to identify sentiments from writing automatically and labor-intensive to manually annotate. If the REALM data is publicly available, it will be great for researchers who are working on opinion analysis.


LiangZhou

---

Comments by ToddDavies

I enjoyed reading this paper. Unfortunately, it looks like the link you give for accessing REALM is no longer active, so I haven't been able to interact with it directly. But your paper gives a good overview and I think I understand roughly what this program does/did.

Your observation that "REALM bypasses the potential for redundancy and obfuscating prose", by encouraging users to identify with and add their tally to arguments that closely match their own sentiments, nicely characterizes what I like most about this software. Although the system appears to be designed for self-reflection and pedagical dialogue, I can imagine implementing something like this in a platform like Deme (our software) that is aimed at deliberation within groups. In our system, a question could be posed for discussion, and the discussion could, perhaps optionally, be conducted in a REALM-like way. This could be especially useful at the early stages when a group is considering a course of action and wants to identify the main pros and cons, for instance.

As a tool for moral reasoning, REALM could be used to expose people to the consequences of their beliefs and arguments. Interesting psychological and neuroimaging studies by Josh Greene and others are showing that people's reactions to moral dilemmas can swing wildly based on extra-rational criteria. For instance, people strongly oppose killing someone in order to save a larger number of people when the act of killing requires them to have physical contact with the person to be killed, but most people switch to a "utilitarian" calculation and approve of the killing when it can be done remotely, e.g. by pulling a switch. These are between-subjects experiments, so participants see only one version of the dilemma. There are many other cases of dramatic effects in, for example, fairness judgments, jury awards, and the like (work by Richard Thaler, Danny Kahneman, etc.). Some of these effects persist within-subjects, and people sometimes hold onto an apparently contradictory judgment even when they are made aware of the contradiction. But at the least, such an exercise should make one doubtful about trusting their own intuitive judgments. Sowing such douts should be an important goal for deliberation, and is difficult to achieve in a systematic way online, where people are less likely to be held accountable for their opinions.

ToddDavies

---

Comments by CliffLampe

So one quick thing on the Slashdot stuff in your introduction. You say “On Slashdot, for example, users frequently give up on threads, posting brain-dump posts at the top level and not bothering to examine what has already been said. This is hardly a debate at all- it is some combination of nitpicking, pontificating and insularity.”

I have to comment on that because just a couple of days ago I attended a Slashdot meeting where they were talking about that exact problem. I don’t think they have a solution picked out (hopefully this workshop provides some good ideas) but some of the things they talked about were reparenting threads that go offtopic, or cutting off threads that have become completely off-topic. In the paper I first authored for the workshop (which was rejected so don’t look for it), I argued that readers who choose flat or nested display modes might be interacting with the discussion differently.

Now for the REALM specific content. I dig it. To some extent, it reminds me of deliberative polling work done by James Fishkin (http://cyber.law.harvard.edu/fishkin.html∞ ) . Fishkin started by having large face-to-face meetings where participants would be asked several questions, then deliberate about the topic, and then polled again. The importance of exposure on persuasion is old hat, but what REALM does is start from the other direction. Rather than deliberating, then polling, REALM polls as a means of deliberation. You can imagine some pretty neat permutations on this, most of which Kushal covers in the Conclusions section. I’d like to add some like connections to digital libraries for evidence-based arguing, and a social presence addition that could be used to get a sense of zeitgeist. Kushal mentions voting in passing, but you could do some pretty interesting things with personalized rating, and group reputation.
I’d like to think some more about how different users will interact. Is social loafing around? One of the things I like about the system is that it constrains trolling, which often is the death of threaded discussion. However, some types of trolling might be beneficial. Does the system constrain user activities too much?

CliffLampe

---

As soon as i started reading the description of the system i thought, "It's Animal!" It's a clever idea. The current tool is a bit restrictive, but i'm excited about the possibilities for extension that you mentioned.

-- KaPingYee