Vaccination on the Internet
The Internet is a frequently used source for health information by the public (Fox, 2006; Fox and Lee, 2000; Harris Interactive, 2002; Liszka et al., 2006). Polls from 2006 show that 80% of Internet users in the United States (Fox, 2006) and 27% of users in Great Britain have sought health-related information online (Pollard, 2006). Health care professionals have expressed concerns about the quality and veracity of information that individuals receive from Internet-based sources (Hardey, 1999; Kunst et al., 2002; Silberg et al., 1997; Walji et al., 2004). One particular area of concern is the use of Internet sites to communicate information to discourage the uptake of routine immunization. Four recent studies examining the quality of information about immunization on the World Wide Web found that search engines returned a high percentage of websites that promoted viewpoints that contradicted conventional medical opinion and advice (Davies et al., 2002; Nasir, 2000; Wolfe et al., 2002; Zimmerman et al., 2005). Davies et al. (2002) found that 43% of websites contained explicit anti-vaccination content and in another study, sites appearing in the top ten results for Internet searches frequently disseminated viewpoints critical of immunization (Nasir, 2000). All four studies found that the Internet is being used by vaccine critics to share health experiences, provide a forum for people who share similar medical opinions, and to disseminate alternative medical viewpoints. For example, Zimmerman et al. (2005) and Wolfe et al. (2002) found common design attributes on vaccine critic's websites indicating social networking, along with marketing features such as links to other anti-vaccination websites. Other features of anti-vaccination Websites included provisions for legal advice for avoiding required immunizations, mechanisms to allow users to donate money to support the community, solicitation of personal stories, active listservs or chatrooms, and opportunities to purchase materials such as books and tapes.
Using YouTube Web Video For Sharing and Communication
Websites offering video streaming to users, such as YouTube (http://www.youtube.com), are becoming increasingly popular and those that allow user tagging, viewer rating, commenting and ranking provide a novel platform for sharing health information and enabling the formation of groups around certain health viewpoints (Cheng et al., 2007; Keelan et al, 2007; Lange, 2007). Sharing health information on YouTube is also becoming increasingly popular (Keelan et al, 2007; Khamsi, 2007). YouTube and other video services therefore have the potential to communicate health information to a large segment of the population. YouTube provides users with a new Internet-based tool to convey powerful images of both the risks and benefits of immunization (Keelan et al, 2007).
Electronic media such as Internet Relay Chat, web-based bulletin boards, and e-mail can be used to connect people together into a social network (Haythornthwaite, 2005). Previous research has examined whether social networks among YouTube users form when users interact with videos that are uploaded to YouTube (Geisler and Burns, 2007; Halvey and Keane, 2007; Lange, 2007). Other researchers have studied the communication patterns within video conversations (Molyneaux et al, 2008a; Molyneaux et al, 2008b). Halvey and Keane (2007) discovered through an analysis of the frequency distributions of views and number of uploaded videos on YouTube, that while many YouTube users did not participate in social networks, they did form social groups. In order to facilitate understanding of the discussion that follows, the terms social network and social group will now be defined. A social network is a graph where the edges between the nodes (typically people) reflect relationships between those nodes. More recently the term “social network” has been co-opted by social networking sites where the social network is a graph constructed by linking all the contact lists of site members together. For instance, if Person B was within the contact list of someone who was in turn in the contact list of Person A, Then Person B would be added to Person A's social network as a “friend of a friend”. Note that in this type of social network, many of the members connected through friends of friends may not know each other, just as many guests at a wedding may not know each other even if they all know the bride or groom. A social group on the other hand, contains people who are linked by a shared purpose and who will generally know each other. Often there may be a group member list and their may be explicit criteria as to who is inside, or outside, the group. While the distinction between communities and groups is somewhat blurred, communities tend to be larger, and they may contain different social groups nested within them. Groups that coalesce around video sharing represent a particular kind of technology enabled group (Lange, 2007).
Social Network Analysis and Cohesive Subgroups
A social network is a graph where the edges between the nodes (typically people) reflect relationships between those nodes. More recently the term “social network” has been co-opted by social networking sites where the social network is a graph constructed by linking all the contact lists of site members together. For instance, if Person B was within the contact list of someone who was in turn in the contact list of Person A, Then Person B would be added to Person A's social network as a “friend of a friend”. Note that in this type of social network, many of the members connected through friends of friends may not know each other, just as many guests at a wedding may not know each other even if they all know the bride or groom. A social group on the other hand, contains people who are linked by a shared purpose and who will generally know each other. Often there may be a group member list and there may be explicit criteria as to who is inside, or outside, the group. Groups that coalesce around video sharing represent a particular kind of technology enabled group (Lange, 2007). Wasserman and Faust (1994) defined a cohesive subgroup as a set of actors (nodes) that are relatively dense and directly connected through reciprocated (bi-directional) relationships (links). Previous research has shown that cohesive subgroups form communities of interest (Dixon, 1981), have weak ties (Garton, Haythornthwaite, and Wellman, 1997), and have cohesive bonds that bring people together (Piper, Marrache, Lacroix, Richardsen, and Jones, 1983).
A variety of methods have been used to identify subgroups in social networks, including in-degree screening for potential subgroups (Kumar, Raghavan, Rajagopalan, and Tomkins, 1999), content analysis of text and tags associated with Web pages (Flake, Lawrence, Giles, and Coetzee, 2002) and threaded conversations (Gruzd and Haythornthwaite, 2008), link analysis (Chau, Shiu, Chan and Chen, 2005), graph theory for finding densely-connected subgraphs within larger graphs (Gibson, Kumar, and Tomkins, 2005), and optimization (Tantipathananandh, Berger-Wolf and Kempe, 2007). However, previous research has tended to involve static networks, web pages as seeds, or links between text content, rather than links between people as reflected in their online interactions.
Social network analysis is a useful method for finding groups in online social networks (Garton, Haythornthwaite, and Wellman, 1997). Clique analysis and related methods look directly at the links that occur in a network and identify specific patterns of connectivity (e.g., subgroups where everyone in the subgroup has a direct connection to everyone else). Cliques and k-plexes have been used to characterize groupings in social networks (Alba, 2003; Balasundaram, Butenko, Hicks, and Sachdeva, 2007; Chin and Chignell, 2007, Du et al, 2007; Reffay and Chanier, 2003; Sterling, 2004; Wasserman and Faust, 1994), where in a clique (Wasserman and Faust, 1994) each member has a direct connection to every other member in the subgroup but in a k-plex, each member in the subgroup has direct ties to at least n-k other members where n is the number of members in the subgroup and k is a parameter. However, finding cliques and k-plexes in large networks is a computationally expensive and exhaustive process that is NP-complete (Balasundaram, Butenko, Hicks, and Sachdeva, 2007), i.e., the computational effort scales exponentially with the number of nodes in the network.
Network centrality (or centrality) (Freeman, 1978) measures how important or central an individual node is to a network. People who are actively involved in one or more subgroups will generally score higher with respect to centrality scores for the corresponding network. Centrality has been used for identifying possible members of cohesive subgroups in networks derived from online interactions (Chin and Chignell, 2007b; Tyler, Wilkinson, and Huberman, 2005; Welser, Gleave, Fisher, and Smith, 2007). Clustering algorithms using centrality measures such as betweenness centrality (Freeman, 1978) have been used for repeatedly dividing the network into clusters and thus automatically identifying cohesive subgroups (Girvan and Newman, 2002; Gloor, Laubacher, Dynes, and Zhao, 2003; Marlow, 2004; Tremayne, Zheng, Lee, and Jeong, 2006; Tyler, Wilkinson, and Huberman, 2005). Betweenness centrality measures the extent to which a node can act as an intermediary or broker to other nodes (Freeman, 1978). In addition, closeness centrality (Crucitti, Latora, and Porta, 2006; Kurdia, Daescu, Ammann, Kakhniashvili, and Goodman, 2007; Ma and Zeng, 2003), and degree centrality (Fisher, 2005; Welser, Gleave, Fisher, and Smith, 2007; Chin and Chignell, 2007b), have also been used to infer community structure and find members of cohesive subgroups. Closeness centrality measures how many steps on average it takes for an individual node to reach every other node in the network, whereas degree centrality measures the number of direct connections that an individual node has to other nodes within a network (Freeman, 1978). Although network centrality measures are easy to calculate using social network analytic programs, there has been no consensus among researchers as to the most meaningful centrality measure to use for finding subgroup members (Costenbader and Valente, 2003).