Article Preview
Top1. Introduction
Deliberation is the act where communities identify possible solutions for a problem and the one(s) from this space that best meet their needs (Walton & Krabbe, 1995; v. Eemeren & Grootendorst, 2003). The spectrum of communities whose discussions rely on reasons and arguments is broad: It not only includes groups of citizens, from (small) municipalities to much larger administrative units. It also ranges from communities in science and technology, including the teams developing software, and communities of online gamers to groups of experts within large companies or organizations. Many communities are small, consisting of about, say, 100 or 200 individuals.
In practice, deliberation faces problems: Major flaws of group discussions are poorly organized content, repetitions, off-topic comments, bad wording and aggressive and conflicting behavior of participants. Some recent projects, e.g., Deliberatorium (Klein, 2011), have tried to apply a very formal argumentation model to bring structure to online discussions and to facilitate content evaluation. However, such rigid formalisms often undermine the natural discussion flow and require a lot of effort from participants. The question we want to investigate here is whether a simple, intuitive argumentation model, but together with ratings by participants, possibly of different type, allows to identify useful points, arguments and convincing proposals.
Designing such a scoring scheme is not obvious. We for our part propose to weight participants based on the adherence to criteria which correspond to efficient discussion behavior, such as the absence of repetition or off-topic comments, clarity of argumentation etc. However, identifying arguments and deriving conclusions and decisions from a discussion still is difficult. Thus, a question we address is to what extent such a derivation can be based on the structure of the discussion. Further, in the discussions foreseen here, there is no objective truth criterion. Instead, criteria we target at include community satisfaction and consensus of opinions. This makes the assessment of approaches such as the one proposed here more difficult. Finally, the broad variety of communities relying on deliberation will make it necessary to accommodate small changes of the scoring scheme, targeting at specific communities. This means that our approach must be robust to such changes.
We have proposed a relatively simple argumentation model to categorize content and different rating types to assess its quality. The rationale has been to give a clear structure to the discussion and to nudge discussants towards deliberation. In more detail, participants discuss different proposals, each one in a separate thread (mainly by posting arguments in favor or against it). Participants also categorize their comments based on its content; examples of respective comment types are ‘pro argument’ or ‘contra argument’. They can also assess comments by other participants, by giving feedback regarding the argumentation presented, post comments that explicitly express agreement or disagreement etc. The assessment can also refer to the clarity of writing, to the tonality of comments, or to the types of the comments. Based on all this information, our approach assesses potential solutions to discussion subjects which participants have proposed in the course of the discussion. With our approach, collecting ideas for solutions is as important as their evaluation. This is in slight contrast to other recent deliberation projects such as ConsiderIt (Kriplean, Morgan, Freelon, Borning, & Bennett, 2012), which focuses on the collection of pro and contra arguments.