Understanding Patterns of Online Discussion

Canberra.
The next speaker at DHA2012 is Sora Park, whose interest is in the processes of online discussion participation, initially especially in the context of the 100 days of political protest in South Korea in 2008. Different online discussion platforms have different affordances, of course – some will list only the most recent or most popular (or most recently popular) posts, for example, thus directing users’ attention towards specific contributions.

Questions around online discussion address topics such as whether there is true debate or just an exchange of partisan statements; whether there is a disparity between readership and authorship; whether the ‘spiral of silence’ which militates against the expression of opposing views is amplified; and whether the increase the polarisation of debates. In reality, of course, the answers to those questions are also dependent on a range of contextual factors: their structural layout; the make-up of the participant base; and the type of topic under discussion.

Within an online community, there are a broad range of roles being played by participants (discussion starters, facilitators, respondents, lurkers, …) – and these roles are not fixed, but also context-dependent. Information flows and influence within a site are also related to these roles – some users emerge as opinion leaders, and relationships emerge between authors, respondents, and readers.

Sora examined this in the context of threaded discussions on opinion sites in South Korea (which operate differently from blogs or social network sites, for example). Questions included which types of posts received the most responses; how people responded to these messages; what patterns of interaction occurred; and whether agreement or disagreement in the discussion affected its flow.

Sora’s focus was on the discussion site Daum Agora, during two weeks in September/October 2008. She tracked some 38,000 posts, and analysed a sample of (eventually) some 580 posts and replies from this dataset (from 85 authors and 314 repliers); the analysis focussed on message topic, content, sources, responses, length, expressions of agreement or disagreement, and the relevance of replies. On average, messages received some 170 views, and just over 4 replies from 2.7 repliers. On average, authors posted 17 messages per week; repliers posted nearly 7 replies per week. Replies were usually considerably shorter than the original posts.

People tended to respond to posts which were relevant to them, and engaged more if the authors responded back again, too. Length of original posts, the amount of information or opinion they contained, and authors’ reputations did not matter much, by contrast. (Longer posts, and messages with proper credits or links, tended to get more relevant replies, though.) If the authors responded again, the discussion unfolded more prominently in a dyadic fashion between authors and repliers than between the repliers. Where agreement was expressed, responses were substantially more likely.

This means there was disparity between authors, repliers, and ‘mere’ readers/lurkers. Authors’ reputations were not a significant factor in interactions; people responded to interesting messages, not recognised authors. Discourse was stronger between authors and individual repliers, rather than between the repliers. Personalised, daily events also received more responses, even in political discussion sites.