What Makes a Successful Online Community?

The first of the afternoon sessions this Thursday at AoIR 2005 is on 'Participation and Trust in Online Communities'. Andrew Cox from the University of Sheffield is the first presenter. 

Andrew Cox: The Parameters of an Online Community

Andrew's work looked at the links and knowledge sharing amongst people working in different organisations and across different organisational jurisdictions - in this case, the Web developers working for universities. There are a number of online spaces available to them, some used, some not, as well as various conferences. in the UK, they continue to use listservs as a key tool, and in the U.S., the main equivalent uwebd is also listserv-based, and has seen a continuing importance (even though we might consider mailing-lists as a somewhat old-fashioned CMC tool by now). There has been some decrease in list membership in recent years, however, declining by some 5% per annum since 2003. It is also interesting that some 25% of list members have turned off mail delivery of the list (presumably accessing the list via the Web?). And there is ongoing churn of members (new members joining, old members dropping out). List traffic is around 120 per month, and again this has declined gradually in recent years.

Andrew investigated a number of aspects in this context. Not many people do post off-list as well (this means that the archive is representative as well), and some 80% of list members are from the academic ac.uk domain; some 20% of members of the central Web teams in UK academic institutions were list members, and there was some considerable (but declining) overlap between list members and conference attendees. List participation was reasonably well spread across list members - the top 13 posters made up for only some 36% of list traffic over 7 years; at the same time, only some 17% of all list members participated in discussions, so there were a large number of lurkers, and there was a strong gender imbalance in favour of male participants.

Andrew also conducted a topic coding exercise for list content, and interviewed participants about their perception of the list. They found it a 'useful source of information', if not critical, and there was some evidence of the emergence of a network of participants (some 50% of respondents said that they had made professional connections). Interactions on the list were relatively brief and mainly in a Q&A genre; there was little narrative or story to the messages.

Some conclusions, then: data about the list is a rich, but under-used source of data; however, the data is relatively fugitive unless it is captured effectively, and there is relatively little baseline data with which to compare this information, as few other studies have been conducted in this line of research.

Anett Kralisch: Risk Reduction Stategies in e-Health

Up next is Anett Kralisch from the Humboldt-Universität Berlin, presenting a paper co-authored with Martin Eisend from the Freie Universität Berlin on risk perception in the context of e-health Websites. Users might use such sites because it is to expensive or embarrassing to speak to a doctor, and they might respond to these risks through risk reduction behaviours, and end up with a sense of satisfaction in their risk reduction strategies. Risk is the importance of potential negative consequences, multiplied by the probability that such negative consequences may occur, and as such is very subjective. The users' cultural background, then, may lead to systematic patterns in risk perception and risk reduction behaviours.

In addition to cultural backgrounds influencing risk reduction strategies, then, we also see the perceived quality of the national health care system influencing users' sense of risk, as well as their own knowledge of the health domain and their expertise in using the Web as having an impact on the risk reduction effort, and their eventual satisfaction with their risk reduction strategies. Anett and Martin's study, then, analysed user behaviour along a number of parameters.

The study findings, however, were very mixed. Cultural factors appeared to have little impact on risk perception; on the other hand, users' cognitive capacities and domain knowledge were the most important factors on risk perception. How to interpret these findings remains open to discussion.

Kjerstin Thompson & Brian Hamman: Factors in Online Community Participation

Kjerstin Thorson and Brian Hamman from the Missouri School of Journalism are next, speaking on participation in online communities. What is it that makes some of these succeed, and others fail? There are a number of key features which may be important here: the frequency of posts (synchronous vs. asynchronous setup, reduced social cues, and response time have an impact here), interpersonal activity (level of relation between individual messages, and presence or absence in patterns of interaction), the presence and role of a moderator (which may avoid flaming, but perhaps also stifle discussion). 

The study then set up a fake university-based Website, RateMyProfessors.com, which allowed the discussion of teaching staff and classes, and experimented with different combinations of frequency, interactivity, and moderation. What they found was that frequency and user attitudes towards the Website were directly related, but interactivity and frequency were reciprocally related (if I parse the graph right). Another experiment used YouthPoliticsAmerica, a moderated site. Interestingly, they found that people were more likely to participate in online discussions when they were moderated (but acknowledge that there are a great number of moderation models which need to be distinguished here).

Ultimately, the outcome here is that different combinations of factors have different effects. Different (social, informational) kinds of communities are suggested by different sets of cues - and personal differences also have an impact here.