Understanding interactions in virtual HIV communities: a social network analysis approach

ABSTRACT

This study investigated the driving mechanism of building interaction ties among the people living with HIV/AIDS in one of the largest virtual HIV communities in China using social network analysis. Specifically, we explained the probability of forming interaction ties with homophily and popularity characteristics. The exponential random graph modeling results showed that members in this community tend to form homophilous ties in terms of shared location and interests. Moreover, we found a tendency away from popularity effect. This suggests that in this community, resources and information were not disproportionally received by a few of members, which could be beneficial to the overall community.

KEYWORDS:

• Weibo HIV community
• social network analysis
• homophily
• popularity effect
• PLWHA interactions

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCiD

Jingyuan Shi http://orcid.org/0000-0001-5675-4512

Notes

1 Given that the 836 members in The HIV/AIDS Weibo Group have joined 2896 other Weibo groups, we categorized these groups according to their themes. First, we randomly selected 300 groups (10.36%) to formulate an exhaustive coding scheme through the open coding method. Weibo groups were inventoried into 14 categories, including (a) HIV/AIDS, (b) homosexuality, (c) health, (d) non-governmental organizations, (e) professional groups, (f) hobbies, (g) fans, (h) parenting, (i) alumni, (j) politics, (k) religions, (l) social, (m) traditional media and (n) miscellaneous (the definitions and examples of each category are provided upon request). After creating the coding scheme, we randomly selected a second set of 300 groups (10.36%) and independently coded them to build inter-coder reliability (Krippendorff’s α = 0.90) (Krippendorff, 2011). Discrepancies were resolved through discussions. Finally, all 2896 Weibo groups were split into four quarters and separately coded by the four authors.

2 The gwidegree stands for a geometric statistic that inversely weighs the value of indegree as a node’s count on statistic increases (Hunter, 2007). This term allows researchers to model the popularity effect in the formation of network ties (Hunter, 2007; Snijders, Pattison, Robins, & Handcock, 2006). A significantly positive coefficient for gwidegree implies that nodes with low indegree are more likely to form a link with those with high indegree in the network (i.e., the presence of popularity effect). Alternatively, a significantly negative coefficient indicates a preference toward the homogeneity of nodes’ indegree. A non-significant coefficient suggests all indegree distributions are equally preferred (Lusher, Koskinen, & Robins, 2013).

3 The interpretation of an ERGM is similar to the interpretation of a logistic regression. The dependent variable of an ERGM is a tie in a network, and the characteristics of network members and network structures are included as independent variables to explain or predict the probability of a tie formation (Robins et al., 2007).