read

References

Citekey: @Gruzd2013

Gruzd, A., & Haythornthwaite, C. (2013). Enabling community through social media. Journal of Medical Internet Research, 15(10). doi:10.2196/jmir.2796

Notes

Highlights

Pescosolido [19] has suggested that a network-centered view of health, based on social network principles, can act as a bridge between medical sciences and individual health experience. Her ideas respond to an increasing recognition of the impact of connectivity and experience: “The individual is seen as embedded in an ongoing relational dynamic with sequences of events seen as patterned, contingent and emergent” (p. 196). (p. 3)

From an analytical perspective, one of the advantages of taking a social network perspective is that the focus is on what people do with each other rather than the medium or face-to-face context through which they do it. (p. 3)

Thus, friendship may be recognized by pairwise exchange of personal information and emotion, discussion of multiple topics, coparticipation in events, frequent interaction, and the use of multiple media. (p. 3)

The basic principles of social network analysis are derived from graph theory and consider actors (eg, people, organizations) as nodes in a network, connected by relations (what they do with each other, eg, provide new information, emotional support, resources, and/or services) that form interpersonal ties. The nature and variety of relations define the kind of relationship between actors, such as an acquaintanceship, friendship, learning, or work relationship. (p. 4)

The configuration of connections is all-important in social networks. These structures show how actors are connected over the whole network, and thus what paths and obstacles there are for contact, information, and resource flow. Among popular aspects considered for networks are the positions of individuals (p. 5)

Our online interactions make these patterns more readily observable, and many examples exist now of how such patterns can be made visible, for example, in social network interaction patterns [31], patterns of text changes in wikis [32,33], and information seeking patterns (eg, Google Flu trends), each of which contributes to understanding emergent community network properties [34]. (p. 6)

One of the main goals of our research is to gain a better understanding of how social media–based information and communication technologies, such as Twitter, enable a distributed group of people to form and maintain an online community. In particular, we are interested in the following research questions regarding #hcsmca: 1. Whataccountsfortherelativelongevityofthisparticularonline community? Is it because of the founder’s leadership and continuing involvement, or are there core members who are actively and persistently involved in this community? 2. Whatisthecompositionofthiscommunityingeneral?And,more specifically, does their professional role determine a person’s centrality within this community? This will allow us to understand generally how professional roles affect online conversational dynamics, and more specifically whether this online community is a welcoming place for a wide range of professionals or is, instead, dominated by professionals from a particular group. (p. 7)

technologies, such as Twitter, enable a distributed group of people to form and maintain an online community. In particular, we are interested in the following research questions regarding #hcsmca: 1. Whataccountsfortherelativelongevityofthisparticularonline community? Is it because of the founder’s leadership and continuing involvement, or are there core members who are actively and persistently involved in this community? 2. Whatisthecompositionofthiscommunityingeneral?And,more specifically, does their professional role determine a person’s centrality within this community? This will allow us to understand generally how professional roles affect online conversational dynamics, and more specifically whether this online community is a welcoming place for a wide range of professionals or is, instead, dominated by professionals from a particular group. (p. 8)

the analysis relied on a type of network called “Name Network” [38]. The Name Network technique examines the content of the messages and connects one person to another if they mention, reply, or repost another person’s tweet [39,40]. (p. 11)

This shows that the #hcsmca community is not fractionated, but rather that participants are all engaged with the single conversational network. (p. 11)

Discovering Community Leaders (p. 12)

A brief examination of the community blog shows that, as expected, the founder of the #hcsmca group is heavily involved in planning and running the community. (p. 13)

Three social network measures were used to locate influential individuals in this community: (1) the total number of messages contributed during the studied period, (2) the number of times a person is mentioned or replied to, that is, their @username is used in a post by someone else (in-degree centrality), and (3) the number of times a person mentions or replies to others, that is, an individual uses another person’s @username in a post (out-degree centrality). (p. 13)

Prestige and In-Degree Centrality (p. 16)

As noted above, the total number of posted messages indicates only the engagement level on the part of an individual rather than the uptake of their contributions by the community. To find out whether personal messages influence others and make them reply or retweet, we examined in-degree centrality (p. 16)

In examining those on this list other than the founder, we noticed that they have something in common. Most of them have a very active online presence in social media in general, not just in this community. (p. 16)

online presence in social media in general, not just in this community. They are also very passionate and active commentators on health matters on Twitter. (p. 17)

In fact, there is a weak, monotonic (nonlinear) positive correlation between the number of followers and the in-degree centrality (Spearman rho=0.23, P<.01). (p. 17)

One possible explanation of this could be that by participating in weekly discussions on #hcsmca, these individuals expose their followers to this community through their tweets on this topic (with the #hcsmca hashtag). (p. 17)

Influence and Out-Degree Centrality (p. 18)

nother group of people who are important within any online community are people who monitor and retweet messages from others. To identify these individuals, we used the out-degree centrality (p. 18)

There is a strong overlap in who is prominent in both the in-degree and out-degree lists. (p. 19)

This suggests a potential method for identifying such accounts in order to exclude them from analyses of social networks: stark differences between in-degree and out-degree centrality may indicate a non-human, or non-community participant within a conversation. (p. 19)

Actor Roles (p. 19)

To address this, we first manually classified each Twitter user in the dataset into one of 11 roles (see Table 4). (p. 20)

We found a statistically significant relationship between professional roles and indegree centrality (explaining about 7% of the variance, P=.003, using 5000 permutations), indicating that some professional groups are more influential in this community. (p. 21)

Next, we attempted to determine which professional groups were more or less likely to influence discourse in this group. Based on the average in-degree centralities for each of the 11 professional groups (see Table 5), social media health content providers were the most influential group with an average in-degree centrality of 2.89. (p. 21)

Another important observation is that although there seems to be a relationship between professional role and in-degree centrality, there is no apparent preferential attachment among people in the same professional group. (p. 22)

This finding was supported by an analysis of variance density test using both the Structural Blockmodel technique (it examines “whether the different classes have significantly different interaction patterns”), and also with the Variable Homophily model (which “assumes that each group or class of actors has a different homophilic tendency” [42]; where homophily is the tendency for connection based on social similarity). Based on this test (run with the 5000 permutations), the professional roles explain only 0.2% of the total variance (P=.005) when run with the Structural Blockmodel and only 0.1% (P<.001) with the Variable Homophily model. (p. 22)

owever, this example has barely covered the beginnings of potential applications. Some key questions that remain and can form the basis of future work are: How do we implement and measure the impact of social media on health for individual patients and for the general population? What single and/or (p. 31)

Blog Logo

Bodong Chen


Published

Image

Crisscross Landscapes

Bodong Chen, University of Minnesota

Back to Home