Social Support in Several Online Communities

For this assignment Beth and I decided to code 20 social support messages, using groups of 5 from 4 different online support groups – anxiety (www.ofear.com), smoking (www.mayoclinic.com), relationship (in particular, infidelity) (www.enotalone.com), and physical abuse (www.healthyplace.com) – so as to get a more representative sample of the internet as a whole. Prior to the experiment we took some time to agree on the definitions of the social support categories to improve our inter-rater reliability. Ultimately we had results of

% inter-rater reliability: 95
Frequency % of messages
Information 17 .85
Tangible assistance 0 0
Esteem support 8 .4
Network support 2 .1
Emotional support 13 .65
Humor 3 .15

Our data revealed a significant amount of information, esteem support, and emotional support. We would expect these high amounts from Walther's four dimensions. The social distance and anonymity aspects suggest that users will be more comfortable responding intimately to people they don't already know, particularly because they can avoid embarrassment, and thus will be more likely to honestly discuss previous experience and provide relevant advice and emotional support. The lack of cues means respondents needn't worry about negative facial expressions or reactions to their openness at the forefront. Additionally, the social distance aspect predicts that the large user base of the internet would lead to available support even in the fairly specific situations we considered, which we found. Finally, the interaction management and access aspects refer to the respondents' ability to take time in reading and responding, and to do so at their own convenience. Thus, we would expect responses to be, in general, more thought-out, relevant and supportive.

Our results partially support the “numbers” theory that the large user base of the internet will lead to decreased noticeability and consequently support. In the smoking group we looked at, there was a huge number of posts, reflecting the prevalence of the issue, and because of this, many posts were left buried and unanswered. However, our other support groups were generally quite responsive – the remaining three topics we looked at were fairly specific in nature, and the community was likely of a smaller size, providing users the ability to command attention and avoid the negative effects of this theory.

Regarding differences compared to Braithwaite, one likely cause for disagreement was simply the nature of the groups considered. Their study considered posts from one support group over the course of one month. Both throughout that month and in the previous time, it is likely that the community and relationships within it developed as people got to know one another – thus we would expect posts that did not serve a “purpose” of containing one or more of these social support concepts. In our case, however, the threads were typically small and the respondents simply responded once, giving a bit of support which generally related to personal experience. Through their experience, respondents typically gained knowledge regarding how to address the issue at hand and were equipped to provide advice, while also being able to relate on an emotional level, explaining our higher results in information, emotional, and esteem support.

The aforementioned nature of the responses also explains our lack of tangible assistance – having never met before, and in most cases not interested in pursuing further conversation (our network support was also very small for this reason) we would not expect respondents to offer tangible favors. This seems like something that would come after time, with more developed relationships - by this argument, the stronger community of the original study could explain their nonzero level of tangible assistance.

The Braithwaite study mentions an idea known as the “optimal matching model,” which could in particular explain our high level of emotional support. The model suggests that emotional support is more likely to be given “when the recipient is experiencing distressful circumstances that are not subject to his or her control.” In our results, the frequency of responses containing emotional support in the infidelity and abuse threads was extremely high, and in both of these cases the original poster was the victim of another's actions, and thus did not have control over the situation, supporting this model.

Humor, though nothing to compare to, was also reasonably small, all things considered. This is likely due to the more serious subject matter of our support groups – only in the case of bizarre social anxieties does it seem jokes might be appropriate, and this was in fact where we found most of our humor.

As a final note, from a statistical standpoint, it is reasonable to also consider the small sample size as a potential reason for discrepancy. While our results may in fact represent true values for support as a whole, which we could justify with the above reasoning, we would need many more messages to have a firm idea of the true percentages (the Braithwaite experiment had almost 1200 messages) which might in fact be closer to their results.