Thesis Proposal

Co-constructing controversy: content analysis of collaborative knowledge negotiation in online communities Noriko Hara and Madelyn Rose Sanfilippo

Department of Information & Library Science, Indiana University Bloomington, Bloomington, IN, USA

ABSTRACT Knowledge sharing online has flourished within organizations as well as open online communities due to the pervasiveness of Web 2.0 platforms. This paper builds on previous studies of social construction of knowledge online and investigates how contributors in online communities collaboratively share and construct controversial scientific knowledge. As the general public participates in such knowledge collaboration, understanding the processes through which they contribute content and roles that they play is imperative. The authors conducted the content analysis of three online communities that engage in knowledge collaboration on the subject of Measles, Mumps, and Rubella (MMR) vaccination, which is perceived as contentious knowledge by the public due to the widespread myth among parents that the MMR vaccine is associated with autism. The study’s findings include that the content discussed is influenced by the purposes of the communities, nature of the tasks, and demographics of participants, although they discussed the same topic. The authors also found that the framework of knowledge reuse and knowledge co-construction sites is useful for investigating the content and roles that appeared in the three communities. The contribution of the paper includes the analytical framework of knowledge reuse and knowledge co-construction, articulation of the content and roles that appeared in online communities, and unboxing of discourses in three different online communities. Future research directions are also discussed.

ARTICLE HISTORY Received 19 February 2015 Accepted 8 January 2016

KEYWORDS Knowledge collaboration; online communities; controversy; Web 2.0

1. Introduction

Knowledge sharing online has recently become pervasive, especially within open online communities (Hara, Shachaf, & Hew, 2010), due to widely accessible Web 2.0 platforms, such as Twitter, Facebook, and Wikipedia (e.g., Flanagin, Hocevar, & Samahitob, 2014). Consequently, ordinary citizens tend to not rely on experts to gain scientific knowledge as in the past, to seek shared knowledge, and to participate in knowledge co-construction in online communities.

Some argue that this may lead to democratization of knowledge (e.g., Qualman, 2012; Tapscott & Williams, 2006). In theory, anyone from any background can share and con- tribute to open online sources such as Wikipedia. Additionally, some scientific knowledge

© 2016 Taylor & Francis

CONTACT Noriko Hara [email protected]

INFORMATION, COMMUNICATION & SOCIETY, 2016 http://dx.doi.org/10.1080/1369118X.2016.1142595

D ow

nl oa

de d

by [

N ew

Y or

k U

ni ve

rs it

y] a

t 06

:4 2

09 F

eb ru

ar y

20 16

is contested even among scientists, requiring ordinary citizens to interpret competing find- ings. In online communities, participants sometimes share contested knowledge and col- laboratively construct understanding of scientific knowledge.

The study described in this paper built on previous work on social construction of knowledge in online environments (e.g., König, 2013) and included more than one kind of community in order to investigate types of interactions and roles played by participants. We selected a contentious scientific topic, child vaccination, which has implications for everyday decision-making. The following question motivated this research: How do people make sense of contested scientific knowledge in online communities? To investi- gate this question, we chose three open online communities that engage in knowledge col- laboration. The specific research questions addressed by this article are: How are independent communities similar or different in terms of the content of knowledge col- laboration?; and how do the roles that participants play in supporting knowledge collab- oration differ or compare across communities?

2. Background: knowledge collaboration

The knowledge collaboration literature (Faraj, Jarvenpaa, & Majchrzak, 2011) frequently discusses two types of activities. One is sharing existing knowledge, sometimes called knowledge reuse (Majchrzak, Cooper, & Neece, 2004; Markus, 2001). Yahoo! Answers represents this type, with participants sharing their own experience and knowledge (e.g., Rechavi & Rafaeli, 2012). The other type includes activities that aim to collaboratively construct knowledge through Web 2.0 platforms. Wikipedia is a salient example of a col- laborative knowledge creation site (Goldspink, 2010; Fichman & Hara, 2014; Kane, Johnson, & Majchrzak, 2014).

Markus (2001) identified four types of knowledge reuse situations: capturing, packa- ging, distributing, and reusing knowledge. Knowledge reuse, as a construct, depends on the assumption that correct knowledge exists and that primary information behaviors are to locate, share, and disseminate. Previous literature examining knowledge reuse has often addressed motivations for sharing, identifying that the objective is for other individ- uals to have access to that knowledge (Hew & Hara, 2007; Nov, 2007). Seebach (2012) along with Hara et al. (2010) classified the type of discourse that knowledge seekers and providers exchange in enterprise microblogging and Wikipedia, respectively. When dealing with knowledge-seeking behavior, some use individuals’ personal networks, for example, through Facebook (Lampe, Vitak, Gray, & Ellison, 2012). Meservy, Jensen, and Fadel (2014) address how knowledge seekers evaluate provided content. They found that the source of expertise and validation that solutions worked in previous cases weigh more in determining the quality of content than actual content quality for the study participants. Judging content is an acute issue on social media such as Twitter (e.g., Liao, Wagner, Pirolli, & Fu, 2012).

Prior scholarship has investigated the collaborative construction of knowledge. The underlying assumption about knowledge in this line of research is that participants debate what acceptable knowledge is and that the goal is to eventually agree on producing representative knowledge. For example, the collection of articles presented in the special issue of Participatory knowledge production 2.0: critical views and experiences (Wyatt, Bier, Harris, & van Heur, 2013) addresses this research domain. König (2013) in that

2 N. HARA AND M. R. SANFILIPPO

D ow

nl oa

de d

by [

N ew

Y or

k U

ni ve

rs it

y] a

t 06

:4 2

09 F

eb ru

ar y

20 16

special issue analyzed the democratization of the knowledge production process in the German Wikipedia. Her examination of the article, ‘the September 11, 2011 attacks in the U.S.’ identified that the existing knowledge hierarchy offline marginalized alternative interpretations of the incident. König asserted that existing knowledge authorities were mirrored in this seemingly open knowledge production site. Similarly, Swarts (2009) studied social construction of knowledge in a Wikipedia entry on clean coal technology. He used Actor Network Theory to analyze the process and specifically focused on two acts: opening moves and closing moves. These are rhetorical moves that shape discourse in terms of presenting ‘facts.’ Swarts revealed the process in which contributors ‘push facts forward, create alliances, hide controversy, and allow collective intelligence to emerge in the form of a stable entry’ (p. 287). Although these studies shed light on how crowds col- laboratively construct knowledge on Wikipedia, it is uncertain what types of knowledge are shared and re-constructed in this process.

Certain knowledge is contentious ‒ there are issues for which no consensus exists among the general public. This includes issues related to science and technology (e.g., climate change; see Jang, 2014) as well as living individuals’ biographies (Joyce, Butler, & Pike, 2011) and health (Kane et al., 2014). Jang (2014) examined four controversial topics in science and identified how individuals’ previous attitudes influence online infor- mation-seeking behavior. As consumers of information, individuals in the study were not greatly influenced by previous attitudes and bias while searching for conflicting knowl- edge. Joyce et al. labeled artifacts or objects as ‘contentious objects’ when conflicts remain unresolved due to differences in groups’ goals. They contended that ‘contentious objects … create affordances for conflict’ (2011, p. 30) and called for more investigation of contentious objects. Kane et al. (2014) conducted another study that deals with a contro- versial topic, that is, autism. They examined the deliberation processes on the Wikipedia discussion forum (called Wikipedia Talk pages) associated with the autism article. They identified three emerging patterns to handle some disputes while contributors engage in the co-construction of the article.

While previous studies are informative, they are lacking in the following aspects: (1) addressing generic topics, but not scientific topics, that affect everyday decisions; (2) focus- ing on one community, commonly Wikipedia; and (3) examining specific content and roles. To address the lacuna in the literature, further examination of content and roles is needed while participants are engaging in knowledge collaboration, especially regarding contentious scientific knowledge. In response, we compare three online communities that deal primarily with collaborative knowledge production in this article.

3. Methodology

Innumerable public controversies have evolved during the social media age, leading to col- laborative and distributed negotiation of truth and perspective in online social environ- ments. While traditional media have long lent particular perspectives to analysis of controversy (Holton, Weberling, Clarke, & Smith, 2012), in recent years, parallel and sub- sequent discussion has continued online (Nagar, 2012). In selecting an appropriate case through which to consider social construction of knowledge, important factors included: multiple distinct points of view, confluence of layperson and scientist perspectives, on- going debate and interest, and social significance of interpretations. These criteria

INFORMATION, COMMUNICATION & SOCIETY 3

D ow

nl oa

de d

by [

N ew

Y or

k U

ni ve

rs it

y] a

t 06

:4 2

09 F

eb ru

ar y

20 16

highlighted a number of possibilities, ultimately leading to the selection of the Measles, Mumps, and Rubella (MMR) vaccine due to the widespread myth among parents that the MMR vaccine is associated with autism. The persistence of this type of vaccine myth was recently reported (Clift & Rizzolo, 2014).

The MMR vaccine became a serious and polarizing controversy as a result of a combi- nation of research misconduct, misinterpretations, media sensationalism, and confusion. Research by Andrew Wakefield posited links between the MMR vaccine and Crohn’s disease as well as Crohn’s disease and autism, yet Wakefield’s research was based on non- random, nonanonymous, and invalid sampling, in addition to being funded by an interest group (Holton et al., 2012; Petts & Niemeyer, 2004). When this work was published, the media publicized it in an unusually aggressive way, and it was publically interpreted as proving autism was caused by vaccination (Holton et al., 2012). Many politicians (e.g., Representative Dan Burton (R-Indiana)) and celebrities, notably Jenny McCarthy who wrote a book on the subject, support this belief (Holton et al., 2012; Petts & Niemeyer, 2004). However, research results were contested and contradicted (Purssell, 2004); many of Wakefield’s co-authors rejected the research, and the journal that published the results later retracted the paper (Petts & Niemeyer, 2004). Yet, in the wake of this con- troversy, fear that the MMR vaccine had led to the autism epidemic spread, and vacci- nation rates declined (Gardner, Davies, McAteer, & Michie, 2010; Shelby & Ernst, 2013), with significant public health impacts due to outbreak and spread of diseases that were virtually absent in developed countries in the previous decade (Gardner et al., 2010). The debate over the MMR vaccine’s risks continues, (Larson et al., 2013) and inter- est in the controversy has grown in parallel with significant events and publicity.

The level of interest in the subject is significant. First, fearful parents of autistic children have latched onto the single study in order to hyper-rationalize autism, and the idea has found support through a combination of social deference to scientific ‘Expertise’ without true understanding (John, 2011). Medical professionals have failed to explain why the fears are irrational and why further scientific investigation through proper procedures invali- dates the Wakefield data and contradicts the findings (Purssell, 2004). Second, the rise in social information sharing (Shelby & Ernst, 2013; Skea, Entwistle, Watt, & Russell, 2008) and sense making (Petts & Niemeyer, 2004) makes analysis of the online commu- nities in which this happens more salient as people from many perspectives co-construct knowledge.

The debate online about the MMR/autism controversy has included parents seeking information from experienced parents, experts seeking to dispel falsehoods and reduce fear, conspiracy theorists, and online knowledge curators, such as Wikipedians and social Q&A users. In this sense, this controversy is ideal for examining contextual differ- ences in collaborative knowledge development because the participants and the commu- nities are so diverse.

3.1. Data

Online communities have extensively debated the relationship between the MMR vaccine and autism, including the Wikipedia Talk community, the Baby Center, and the Berkeley Parents Network. They represent very different combinations of pro- and anti-vaccine perspectives, along with concerned parents interested in the truth, general conspiracy

4 N. HARA AND M. R. SANFILIPPO

D ow

nl oa

de d

by [

N ew

Y or

k U

ni ve

rs it

y] a

t 06

:4 2

09 F

eb ru

ar y

20 16

theorists, and representatives of the public health and medical communities. Furthermore, the different patterns of behaviors and uses of information can lead to analysis along two comparative lines: co-construction of knowledge, exemplified by Wikipedia, and re-use of knowledge, exemplified by the Baby Center and the Berkeley Parents Network. It is impor- tant to examine multiple communities as they seek truth and understanding about the MMR vaccine and autism because a comparative perspective illustrates the impact, if any, of context, as well as the sources of difference in social understanding of the contro- versy (see Table 1 for a summary of the data sources).

3.1.1. Wikipedia Considerable attention has been paid to understanding Wikipedia as an online commu- nity, an information resource, and a platform for collective sense making (Nagar, 2012). While the Wikipedia articles receive substantial amounts of public attention, Wikipedia’s Talk pages, in which contributors discuss the content of articles, are rich sources for the process of knowledge co-construction. Wikipedia Talk pages are attached to individual articles and archive all of the discussions that occur regarding the article including the content, editorial decisions, and behaviors of contributors. As a venue for interpretation and representation of collective knowledge, its strict policies institutionalize community expectations for practice to ensure that sense making is not co-opted or subverted by poor practices (Nagar, 2012) or deviants (Shachaf & Hara, 2010). Providing protocol for behavior and mediation and intervention in the case of noncompliance is significant to collaborative processes for knowledge co-construction surrounding contentious, hotly debated topics.

It is within this complex context that Wikipedians’ sense-making process seeks a neutral point of view between all facts and interpretations within the community. Looking specifically at the sub-community that will be analyzed, collaborators in con- structing knowledge seek to represent the MMR vaccine, and its polarizing position in society, to provide a public reference. The community seeks less to debate the issue and more to ensure balanced coverage, though individuals sometimes detract. Contributors came from primarily the USA, the UK, and Australia, most of whom follow community policies and norms. A few contributors have dominated particular threads at particular times. The community of contributors was most robust and communicative when the page began in 2007, though many contributors still participate in similar pages.1

3.1.2. The Baby Center Thirty-six million mothers worldwide reference the Baby Center through 14 distinct web- sites in 11 different languages (Promisel & Centinaro, 2014). It is administered by Johnson & Johnson (Promisel& Centinaro, 2014), which has a prominent legacyin products

Table 1. Data sources and distribution. Baby Center Berkeley Parents Network Wikipedia

Criteria for inclusion

All references to MMR and/or MMR controversy

All references to MMR and/or MMR controversy

MMR talk page; MMR vaccine controversy talk page

Posts 541 98 705 Threads 33 18 119 Coverage 2007–2013 1999–2006 2004–2013

INFORMATION, COMMUNICATION & SOCIETY 5

D ow

nl oa

de d

by [

N ew

Y or

k U

ni ve

rs it

y] a

t 06

:4 2

09 F

eb ru

ar y

20 16

for children and promoting informed parenting. Focusing specifically on the US-targeted website, the Baby Center involves mainstream discussions on parenting, including questions for and feedback from experts. Sub-communities have formed around seekers and sharers of health information, with extensive discussion on vaccination and significant debate over the MMR vaccine in particular. There is a highly visible group of contributors who identify as vaccine skeptical. Whether they are a minority or majority, they dominate these debates. The discussions often appeal on an emotional level and are active diachronically.

This community has been subject to some research, mostly with emphasis on content rather than knowledge collaboration processes. The Baby Center has been evaluated with respect to: trust production and privacy concerns in online social exchange (Luo, 2002); the role of site features, user attributes, and sponsor reputation in impacting perceived information credibility (Flanagin & Metzger, 2007); and flexible personalization of online services to users (Borenstein & Saloner, 2001). Scholars have determined that the Baby Center provides important services to parents and caregivers, which are more specifi- cally targeted and approachable than offline parenting information equivalents. However, it remains to be determined how useful and reliable the information from this community is, particularly when the status of ‘truth’ is contested within the community and other communities.

3.1.3. The Berkeley Parents Network The Berkeley Parents Network is an online forum that includes parents primarily from a closely proximate region in the San Francisco Bay area, with around 33,156 members who engage in parent-to-parent exchanges and receive weekly newsletters (Berkeley Parents Network, 2014). This network was founded in 1993 to support parents who worked or were students at the University of California, Berkeley, but has expanded to other parents in the area (2014).

Discussions stem from personal questions to guide decision-making, rather than an interest in unprompted knowledge sharing or general discussion. Threads are often short, ending when parents have been answered. The community favors progressive per- spectives and approaches, as indicated by the number of threads on alternative medicine. Discussions on MMR vaccine concerns reflected individual experiences, such as with aller- gies, and circumstances, somewhat independently of the general controversy, and ceased in 2006, before the controversy peaked in other communities and popular media. This forum has not been studied, though it appears similar to other online forums, albeit with approximate geographic boundaries encompassing a majority of participants.

All complete, publicly available threads pertaining to or including discussion of the MMR vaccine and the vaccine-autism controversies up until 15 August 2013 were har- vested from each of these three communities. This included 705 talk posts from Wikipe- dia, 541 posts from the Baby Center, and 98 posts from the Berkeley Parents Network. Data were collected and considered with posts as the unit of analysis, except for analyzing discussions to distinguish controversial from noncontroversial threads.

3.2. Analysis

Consideration of these posts and threads through textual analysis was systematic. First, a codebook was developed to apply to social construction of knowledge and social

6 N. HARA AND M. R. SANFILIPPO

D ow

nl oa

de d

by [

N ew

Y or

k U

ni ve

rs it

y] a

t 06

:4 2

09 F

eb ru

ar y

20 16

information sharing in a previous study about the Fukushima disaster with Wikipedia (Hara & Doney, 2015). The codebook was revised, so as to specifically relate to the types of online communities addressed in this study. Table 2 presents the definitions of all codes applied, with the exception of dates, which were also applied to reveal longitudi- nal patterns and instances of correspondence between bursts of collaborative activity and important events along the historical timeline of controversy development.

Codes were applied as appropriate once per post, without limit to the total number of codes that could be applied to a post. Frequencies, as the percentage of total posts in a specific community that exemplified particular themes, were then compared across com- munities. Codes were applied by the investigators and by an additional coder. Further- more, qualitative assessment of threads, assessed as posts, were coded in thread sequence, allowing deeper understanding of the nature of communities and the status of understanding.

Second, we coded the posts based on roles that each participant played. A codebook was developed based on the literature review (see Table 3). The roles appear in three different contexts: one for the roles that appear while participants engage in identifying and sharing knowledge, the second for the roles that modify shared knowledge, and the third for the

Table 2. Codebook for content. Context Code Definition

Editorial Editorial Comments related to the editorial aspects of the article; for example, the length of the article, naming of categories, etc.

Knowledge components

Background knowledge

Some background knowledge about peripheral topics; for example, autism generally, availability of the MMR vaccine, other vaccines, vaccine preservatives, or the measles disease

Definitions Definitions of terminology related to the topic Facts about the topic Claimed facts about the topic; for example, details about the Wakefield

study and subsequent public fallout; discussions about the controversial and contested link between the MMR vaccine and autism as well as the MMR vaccine itself.

Media report about the topic

Reporting of media coverage about this event; for example, USA Today reported …

Resources Sharing of any useful resources related to the topic, but not necessarily related to facts per se

Knowledge scaffolding

Citations Any discussions of or citing of sources to justify facts or background knowledge for the purpose of convincing others

Claimed expertise This refers to users who base their arguments on their educational or professional expertise (i.e., ‘as a pediatric RN’) or asks for the expert credentials of other users or resources. This also includes any instances in which participants explicitly claimed no expertise.

Confusion This refers to a user either identifying confusion in other users’ arguments or explicitly stating that they are confused by the facts or presenting contradictory information in the same post

Credibility Instances in which users evaluate the credibility of another actor’s claims, including their arguments, relevance of expertise, etc.

Experience Presents information about their own lived experiences and social knowledge from their offline lives, associated with the topic

Inquiry This refers to actual questions posed to stimulate discussion Interpretation or opinion

This refers to associated discussions in which the poster restates their opinion of another user’s comments or says ‘in my opinion … ’ or presents information on the topic or as background information and then explains how they will act on that or what it means to them. This also includes instances in which underlying motivation, values, or logic was discussed with respect to facts or background information as evidence of interpretation.

Vandalism Vandalism Discussions about how to deal with malicious users

INFORMATION, COMMUNICATION & SOCIETY 7

D ow

nl oa

de d

by [

N ew

Y or

k U

ni ve

rs it

y] a

t 06

:4 2

09 F

eb ru

ar y

20 16

roles that facilitate knowledge-sharing discourses (Hara & Sanfilippo, in preparation). Fur- thermore, we added two more roles based on the data: distractor to the first category and judge to the second category. Distractors seek to have a negative impact on the process either by interrupting, trolling, or vandalizing. These users often include promoters seeking to advertise a semi-related product, service, or alternate information source, such as personal blogs, by spamming the thread. Distraction also occurs when those who play constructive roles engage with distractors and thus digress from the discussion subject. Discussants are not necessarily distractors when they simply hold false or unpop- ular opinions. Judges serve to evaluate and/or demand justification for changes in content, process, and style. Judges also evaluate personal statements and actions, though they do not judge individuals as people. Inter-rater reliability calculated using Cohen’s κ for both roles and contexts between two coders range from 0.43 to 1 (see the Appendix). Hallgren (2012) delineated that 0.41–0.60 indicate moderate agreement, 0.61–0.80 mean substantial agreement, and 0.81–1.0 would be interpreted as almost perfect agreement.

Third, the data were aggregated based on the type of sites, that is, knowledge reuse versus knowledge re-construction sites. For the content, we aggregated the data based on the purposes of contents into four categories: knowledge components, knowledge scaf- folding, editorial, and vandalism. In addition, we created aggregated frequency rows based on three types of roles: identifying and sharing knowledge, modifying knowledge, and facilitating knowledge collaboration. Then, cross-tabulation analysis was conducted using the statistical package R (version 3.0.1).

4. Findings

4.1. RQ 1: how are three communities similar or different in terms of the content of knowledge collaboration?

Generally, all three communities have interpretation/opinion and background knowledge among the top three codes (see Table 4). The communities categorized as knowledge reuse have the same top three codes: interpretation/opinion, background knowledge, and experience. This is probably due to the fact that knowledge reuse communities take advan- tage of their own experience, opinions, and existing knowledge to contribute to the discus- sions. In contrast, within Wikipedia, a knowledge co-construction site, editorial discussions become more prominent as the users collaboratively create and edit an article. In addition, considering the nature of knowledge co-construction sites used by the general public, Wikipedia suffered from vandalism more than the other two commu- nities. When we examined the defining characteristics of each community separately, differences between the communities also became evident.

With respect to Wikipedia, the top three content categories were interpretation/ opinion, background knowledge, and editorial (80.14, 40.71, and 35.18%, respectively). There were considerable personal discussions and arguments unrelated to the content, with extended personal attacks and discussion of interests adversely impacting attempts to write and maintain a neutral point of view. These tangents and various edit wars led to multiple interventions by Wikipedia; interestingly, ‘Jimbo Wales,’ the Wikipedia founder, served as a moderator for the talk page for MMR vaccine controversy at a con- tentious point. Debate over the appropriateness of including quotations from famous

8 N. HARA AND M. R. SANFILIPPO

D ow

nl oa

de d

by [

N ew

Y or

k U

ni ve

rs it

y] a

t 06

:4 2

09 F

eb ru

ar y

20 16

Table 3. Codebook for participant roles. Context Role Definition

Identifying and sharing knowledge

Distractor Those who play this role often seek to have a negative impact on the process, either by distracting, interrupting, trolling, or drive-by vandalism.

Help-giving [giver] (Hara et al., 2010; Hargadon & Bechky, 2006; Kane et al., 2014)

Helpers specifically provide information, knowledge, and experience related to the subject of the thread or content of the knowledge being constructed, based on help sought by the original poster or any mover who reshaped the overall discussion.

Help-seeking [seeker] (Hara et al., 2010; Hargadon & Bechky, 2006; Kane et al., 2014)

Hargadon and Bechky assume a process-based approach in which help is sought and given, with a collective sense- making approach.

Mover (Swarts, 2009) Opening and closing movers serve as highly specialized facilitators by transitioning the discussion to new or adjacent points, as well as funnelling the discussion to specific aspects.

Modifying knowledge

Judge Judges serve to evaluate and/or demand documentation or justification for changes in content, process, and style. Judges also evaluate personal statements and actions, though stop short of judging individuals as people.

Knowledge shaper (Majchrzak, Wagner, & Yates, 2013)

Shapers often remove content and rewrite, without adding new content, in order to convey coherence or particular perspectives. This role simplifies content, removing repetition, and makes information accessible.

Organizer (Faraj et al., 2011) This type of actor synthesizes and represents the information shared by other users into a more usable set of links or facts. Organizers summarize discussions, rather than shaping content.

Reflective reframing [reframer] (Hargadon & Bechky, 2006; Kane et al., 2014)

This role seeks to build on the assumed process-based approach in which help is sought and given, with a collective sense-making approach to evaluating help; reframers reconstruct the consensus, serving to translate information collectively deemed useful from conversations into useful and consistent content.

Reflective reinforcing [reinforcer] (Hargadon & Bechky, 2006; Kane et al., 2014)

Individuals playing this role seek to reinforce identified truths; reinforcers reconstruct the consensus, serving to repeat conclusions and provide evidence in support of consensus.

Facilitating knowledge collaboration

Cross-thread connectivity [connector] (Kuk, 2006; Kane et al., 2014)

Some individuals are interested in issues that weave across threads, leading them to assume a role in which they coordinate multiple topics to provide consistency and continuity across online communities.

Facilitator (Kane et al., 2014) Facilitators assume the role of smoothing and expediting the production process within ongoing discussions by seeking to identify quorums and moderate discussions that will lead to progress.

Governance-oriented approach [governor] (Schroeder & Wagner, 2012)

Experienced users and those with an interest in protecting the status quo employ institutionalisms to prevent new comers from making drastic changes from process or to content.

Mediator (Faraj et al., 2011) Mediators step in when tensions run high within a community and when there are entrenched interests in multiple perspectives. They are similar to facilitators, but they specifically manage conflicts within discussions.

Supporter (Faraj et al., 2011) This type of user both provides support for particular positions articulated within discussion and seeks out references and resources to support claims and positions. Supporters generally do not contribute to content or provide new arguments within discussions.

Unmasker (Faraj et al., 2011) When anonymity or identity deception is allowing an actor to harm a community, an individual might seek to become an unmasker by investigating the behaviors of possible perpetrators in order to either expose the harmful actor or force them to apologize.

INFORMATION, COMMUNICATION & SOCIETY 9

D ow

nl oa

de d

by [

N ew

Y or

k U

ni ve

rs it

y] a

t 06

:4 2

09 F

eb ru

ar y

20 16

people as evidence within the article had reached a stalemate; Jim Wales offered the fol- lowing comments, diffusing the situation:

Although I think Bill Gates is likely correct here, I agree that his views don’t need to be pre- sented in this article. He is a major philanthropist in the area of vaccines, and as is his intense nature as a person, he seems to have done very intensive study, so I think he knows what he is talking about. However, rather than an off the cuff remark from a celebrity businessman/phi- lanthropist style interview, it would be better to stick to verifiable medical sources. ‒ Jimbo Wales (talk) 09:38, 8 February 2011 (UTC)

Therewas also significantdebate over credibility ofusers (31.91%), citations (30.78%), and validity justifications for individual opinions (80.14%). For example, users debated whether the controversy article should be expanded to include other controversies surrounding the MMR vaccine, such as vaccine overload, concerns about the Urabe mumps strain within the vaccine, and associations to other vaccine controversies. In shaping a narrative about the MMR vaccine and controversy, issues of validity and bias dominated user concerns.

A theme that recurred within the Baby Center was emotionalism. Specifically, while instances of interpretation/opinion codes in other communities indicated what the partici- pants took from the facts and background knowledge and how this might impact actions, Baby Center participants emphasized feelings and beliefs. An example of an emotion- driven inquiry is as follows:

i just took my son Kaston to his 2 month checkup and they informed me that people are in mass confusion that this shot could cause autism … im a 1st time mommy and now that i have read some things about it, the statistics of 1 in 150 could be affected by this shot!! what the hell!! are you serious!!! im scared to death to get his shots, but im also scared to death that he may get sick if i DONT!! if anyone could please help me with some info on wether i should or should not get his shots, i would greatly appreciate it!! Posted: 10/26/2011 by ladyethinol

There was particular emphasis on trust and distrust, tying this code closely to credibility (22.55%) in many instances, as well as evidence of fear more than skepticism. The discus- sions tended to stray both from the original post topic and the specific MMR/vaccine con- troversy into general autism and vaccine discussions, with participants providing significantly more background knowledge and experience than facts. Many participants seemed to value experience over both scholarly research and media or celebrity narratives.

Table 4. Distribution of content codes as percentages of posts within communities. Codes Baby Center Berkeley Parents Network Wikipedia

Background knowledge 60.44 74.49 40.71 Citations 10.35 12.24 30.78 Claimed expertise 8.50 12.24 1.84 Confusion 9.24 7.14 8.09 Credibility 22.55 8.16 31.91 Definitions 2.22 0.00 3.12 Editorial 0.00 0.00 35.18 Experience 41.40 62.24 2.55 Facts 14.42 19.39 22.13 Inquiry 16.27 24.49 14.33 Interpretation/opinion 80.41 57.14 80.14 Media reports 2.96 9.18 5.82 Resources 21.63 24.49 13.33 Vandalism 1.29 0.00 4.68

10 N. HARA AND M. R. SANFILIPPO

D ow

nl oa

de d

by [

N ew

Y or

k U

ni ve

rs it

y] a

t 06

:4 2

09 F

eb ru

ar y

20 16

Discussions within the Berkeley Parents Network spanned a much greater breadth than within the other communities; specifically, even when threads began with the MMR vaccine controversy, posts considered public health consequences of nonvaccination, allergies, doctor–patient interactions, and policies. The discussants largely observed a question and answer structure, and when a participant strayed toward debate, the others re-identified and redirected the discussion back toward the original inquiry. This anonymous contribution illustrates an example of redirection:

Please listen to your doctor! The question is not ‘When was the last time anyone got measles around here?’ It should be ‘How soon before an outbreak of this serious disease occurs because increasing numbers of well meaning (but misguided) parents defer or avoid MMR vaccinations for their children?’ … We can’t all be experts on science or statistical research, but given that fact, we need to follow the current advice of a pediatrician rather than be guided by our own uninformed feelings or suspicions. Remember, this is not a personal choice issue ‒ your decision affects the whole population. For the record, my son has received the MMR vaccination, and no, I’m not a doctor.

Opinions were valued far less often than experiences, and media coverage was acknowl- edged as having prompted many of the inquiries, rather than employed as support for facts or arguments. This was particularly evident when users sought to know how other parents had handled immunization issues with their own children; users did not seek opinions on what they should do, but rather exemplar experiences. The scope of discussion was largely personal, yet it was rational rather than emotional compared to Baby Center in terms of the arguments made and the language employed.

Table 5 presents the 14 codes of content aggregated into four categories: knowledge components, knowledge scaffolding, editorial, and vandalism between knowledge reuse sites (Baby Center and Berkeley Parents Network) and knowledge co-construction site (Wikipedia). A χ2 of the content categories, excluding vandalism, indicated a statistically significant difference between the knowledge reuse and knowledge co-construction sites (χ2 = 12.9982, df = 2, p = .0015). The descriptive data showed the major differences between the knowledge reuse and co-construction sites were editorial and vandalism. Even though the value for the vandalism is small, when included in the χ2 test, the results indicated that the difference between knowledge reuse and knowledge co-construc- tion sites was statistically significant (χ2 = 13.756, df = 3, p = .0033).

4.2. RQ 2: how are the three communities similar or different in terms of the roles that participants play in supporting knowledge collaboration?

In examining roles, as was expected, all three communities have giver as one of the top three most common roles (see Table 6). The communities categorized as knowledge

Table 5. Cross-tabulation results of aggregated content and types of sites in percentage. Knowledge co-construction

site (Wikipedia) Knowledge reuse sites

(Baby Center & Berkeley Parents Network)

Editorial 11.94 0 Knowledge components 28.88 35.84 Knowledge scaffolding 57.58 63.78 Vandalism 1.58 0.37

INFORMATION, COMMUNICATION & SOCIETY 11

D ow

nl oa

de d

by [

N ew

Y or

k U

ni ve

rs it

y] a

t 06

:4 2

09 F

eb ru

ar y

20 16

reuse both have giver as most common and supporter as third most common. Since the purpose of knowledge reuse sites is sharing and reconstructing the knowledge, it is natural that giving occurs most frequently. The supporting role helps sustain a collegial community and move the discussion into consensus. These two communities also did not have the role of governor, as they do not have extensive rules for using the sites, unlike Wikipedia.

On the collaborative knowledge creation site, Wikipedia, judge is the most frequently observed role. This is counterintuitive because the role of judge does not produce content. However, the role a judge plays is similar to the knowledge shaper described in Majchrzak et al. (2013). For example, a post commenting on a citation was coded as judge:

There’s a big difference between being a key figure personally involved (like Horton, who approved the original article for publication) plus having written a book on it, and having expressed an opinion on the controversy. ‒ Jkpjkp 05:45, 30 August 2007 (UTC)

Their contribution to knowledge collaboration is significant because it evaluates the content’s value, for example, whether part of content is accurate or whether a sentence should be deleted. It plays a role similar to metacognition in an individual’s mind. In this community, all of the roles happened in the discussions. The least occurring role was unmasker (0.43%).

Two communities, Baby Center and Wikipedia, have the role of distractor as the second most frequent role. When participants engage in controversial topics, they have strong opinions about specific interpretation of these topics. The topic examined here, child vac- cination, is a controversial topic. As a result, we sometimes see digressions within discus- sions by people who have an agenda different from knowledge reuse or knowledge co- construction. These individuals are not contributing to the discussions, but rather divert- ing the discussions.

Despite the fact that the participants engage in the same controversial topic, the Berke- ley Parents Network community did not have substantial appearances by distractors. Instead, seekers and reframers played major roles in this community. This may be due to the fact that participants here tend to seek out progressive perspectives and approaches. In addition, this community is smaller and less well known than the other two. As a result,

Table 6. Distribution of roles as percentage of posts within community. Codes Baby Center Berkeley Parents Network Wikipedia

Connector 2.03 3.06 5.82 Distractor 28.47 8.16 24.68 Facilitator 1.85 0.00 1.28 Giver 59.52 67.35 17.30 Governor 0.00 0.00 8.09 Judge 11.83 2.04 34.61 Knowledge shaper 0.55 1.02 7.94 Mediator 4.44 3.06 3.69 Mover 1.48 7.14 10.92 Organizer 0.92 1.02 2.13 Reframer 5.73 18.37 7.66 Reinforcer 2.96 9.18 5.96 Seeker 9.98 22.45 7.66 Supporter 12.94 18.37 14.61 Unmasker 1.11 0.00 0.43

12 N. HARA AND M. R. SANFILIPPO

D ow

nl oa

de d

by [

N ew

Y or

k U

ni ve

rs it

y] a

t 06

:4 2

09 F

eb ru

ar y

20 16

discussions are inclined to stay focused on the topic specified by the initiator of a given thread. Within the Berkeley Parents Network, two additional roles also did not appear: facilitator and unmasker. Such roles likely were unnecessary because this community had fewer and less significant disputes. When disagreements occurred within controversial threads, the majority of participants ignored distractors or sought to return the discussion back to the topic, thereby playing the role of reframer.

Within the Berkeley Parents Network, givers dominate the discussions, providing the majority of the content in threads, as well as the majority of the knowledge being co-con- structed. Seekers play the second most numerous role, closely followed by the aforemen- tioned reframers and supporters. The dynamic of conversations within this community is much more oriented toward reuse of knowledge than on consensus forming, significantly impacting the distribution of roles.

Table 7 represents aggregated codes of the three categories for the 15 roles between knowledge reuse sites (Baby Center and Berkeley Parents Network) and knowledge co- construction site (Wikipedia). A χ2 indicated that there was a statistically significant differ- ence between the knowledge reuse and knowledge co-construction sites (χ2 = 9.2758, df = 2, p = .0097). The descriptive data indicated that the majority of roles appearing in the knowledge reuse sites were identifying and sharing knowledge roles, whereas roles are spread among the three in the knowledge co-construction site.

5. Discussion and conclusion

In this paper, we addressed the research question: How do ordinary citizens make sense of contested scientific knowledge in online communities? We analyzed three different open online communities ‒ two focused on knowledge reuse and one on knowledge co-con- struction. We found that the three communities examined in the study differed in terms of the content that they discussed, but also that the knowledge reuse sites shared similar content. Likewise, the roles that participants play in these communities are differ- ent, but knowledge reuse sites show some resemblance to one another. The cross-tabula- tion analysis showed statistically significant differences between knowledge reuse and knowledge co-constructions sites in terms of content and roles.

Although the roles that appear in online communities have been discussed in prior lit- erature, they have not been explicitly defined. One of the contributions of this paper is the articulation of these roles based on the literature. Moreover, we identified two additional roles that play important functions in online communities that engage in knowledge col- laboration: distractor and judge. Although distractors do not necessarily make productive contributions to knowledge collaboration, their roles affect identification and sharing of knowledge. Distractors are not likely to appear in closed online communities, such as those in corporate sites because of the expectations to act professionally. However, the

Table 7. Cross-tabulation results of aggregated roles and types of sites in percentage. Knowledge co-construction

site (Wikipedia) Knowledge reuse sites

(Baby Center & Berkeley Parents Network)

Identify and sharing 39.65 68.48 Shaping 15.51 8.97 Supporting 22.19 15.49

INFORMATION, COMMUNICATION & SOCIETY 13

D ow

nl oa

de d

by [

N ew

Y or

k U

ni ve

rs it

y] a

t 06

:4 2

09 F

eb ru

ar y

20 16

distractive roles, such as those played by trolls, are not uncommon (see e.g., Shachaf & Hara, 2010). As we crowd-source knowledge collaboration (Wiggins & Crowston, 2011), managers and site owners need to pay attention to distractors and find a way to cope with them. Joyce, Pike, and Bulter (2013) proposed the strategies of managing con- tentious objects that have implications for how to manage distractors. However, we need to explicitly examine the strategies for dealing with distractors. The other role that we ident- ified was judge. Those who play the role of judge contribute to shaping of knowledge in general. This role appeared to be more important in knowledge co-construction sites, as one-third of the participants in Wikipedia are coded as playing a judge role. When par- ticipants are socially constructing ‘knowledge’ online, someone needs to evaluate and make judgments about the quality of contributed content. We argue that this role is extre- mely important, though it has not been clearly addressed in the previous knowledge col- laboration literature, despite the emphasis on the importance of knowledge shapers (e.g., Majchrzak et al., 2013; Yates, Wagner, & Majchrzak, 2010).

In examining three online communities, we articulated a framework to examine the content of discussions in depth. We identified four major content categories (i.e., editorial, knowledge components, knowledge scaffolding, and vandalism) in these communities that engage in knowledge collaboration activities. Content analysis laid the groundwork to help us understand interactions happening in these sites. Other researchers can use this frame- work to analyze and compare other topics as well as other sites.

Although this study examined the discussions of a specific vaccine called MMR, the implications of the study expand beyond this vaccination and include other scientific con- troversial topics such as climate change or other types of vaccinations. This study sheds light on the practice of participating in knowledge co-construction. As more people par- ticipate in this type of knowledge production, it is important to reveal such practices because this will help scientists and policy makers understand how the general public com- prehends controversial scientific knowledge. Scientists and policy makers can adjust their practices and policies accordingly in terms of how to educate the general public about these topics.

One limitation of the study was that it relied on a single contentious topic, that is, MMR vaccination. Second, because of the popularity of the topic, the discussions may not be considered typical. Third, we chose only three online communities. Although the numbers of posts in each community were not insignificant, if we had added more com- munities to the sample, findings might be different. At the same time, the tradeoff for including more than three would have meant that we could not conduct detailed analysis. For this reason, our limitation of the number of communities evaluated was justified.

Despite these limitations, we believe that this study makes important contributions to the field. First, we examined three online communities and found that the distinction between knowledge reuse sites and knowledge co-construction to be useful. We also articulated roles played in knowledge collaboration sites and identified new roles that have not been well examined in the knowledge collaboration literature. Similarly, we also identified different types of content appearing in these online communities. The con- tribution of the paper includes the analytical framework of knowledge reuse and knowl- edge co-construction, articulation of the roles that appeared in online communities, and unboxing of discourses in three different online communities.

14 N. HARA AND M. R. SANFILIPPO

D ow

nl oa

de d

by [

N ew

Y or

k U

ni ve

rs it

y] a

t 06

:4 2

09 F

eb ru

ar y

20 16

In the future, we can expand the sample to include more open online communities. We are currently conducting a study to examine how the different roles play in controversial and noncontroversial threads. Another study underway is investigating the sequence of roles in threads. In addition, some roles that we identified can be investigated further, such as that of judge, as Majchrzak et al. (2013) scrutinized the role of knowledge shaper. The interplay between the roles and contents that we identified can be analyzed. We can also expand the study to include different controversial topics in science. This study shed light on how interactions occur in three online communities that engage in knowledge collaboration, in a case of contentious scientific knowledge. There is no ques- tion that more studies are needed in this area of research.

Note

1. There is significant overlap in contribution between the pages for MMR vaccine and pages for: Varicella, Vaccination, Hives, Reye’s Syndrome, Autism spectrum, MMRV vaccine, and Vari- cella vaccine. Pages with common contributors to the MMR vaccine Controversy page include: Autism spectrum, The Lancet, BMJ, and Andrew Wakefield, among others. This was measured by considering contributors as edges between pages, as nodes, using NodeXL.

Acknowledgements

We are greatly appreciative of Kjersten Hild and Jylisa Doney for her help in coding sections of the data to establish inter-rater reliability.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Noriko Hara is an associate professor in the Department of Information & Library Science, School of Informatics and Computing at Indiana University. Her research in Social Informatics empha- sizes online knowledge collaboration, communities of practice, and collective behaviors in mediated environments. [email: [email protected]]

Madelyn Sanfilippo is a doctoral candidate in Information Science at Indiana University, Bloo- mington’s School of Informatics and Computing. Madelyn is specifically interested in social and political issues surroudning information and information technology access. [email: mhomuth@ indiana.edu]

References

Berkeley Parents Network. (2014). About the Berkeley Parents Network. Retrieved from http:// parents.berkeley.edu/FAQ/about.html

Borenstein, S., & Saloner, G. (2001). Economics and electronic commerce. The Journal of Economic Perspectives, 15(1), 3–12.

Clift, K., & Rizzolo, D. (2014). Vaccine myths and misconceptions. Journal of the American Academy of Physician Assistants, 27(8), 21–25.

Faraj, S., Jarvenpaa, S., & Majchrzak, A. (2011). Knowledge collaboration in online communities. Organization Science, 22(5), 1224–1239.

INFORMATION, COMMUNICATION & SOCIETY 15

D ow

nl oa

de d

by [

N ew

Y or

k U

ni ve

rs it

y] a

t 06

:4 2

09 F

eb ru

ar y

20 16

Fichman, P., & Hara, N. (2014). Global Wikipedia. International and cross-cultural issues in online collaboration. Lanham, MD: Rowman & Littlefield.

Flanagin, A. J., Hocevar, K. P., & Samahitob, S. N. (2014). Connecting with the user-generated Web: How group identification impacts online information sharing and evaluation. Information, Communication & Society, 17(6), 683–694.

Flanagin, A. J., & Metzger, M. J. (2007). The role of site features, user attributes, and information verification behaviors on the perceived credibility of web-based information. New Media & Society, 9(2), 319–342.

Gardner, B., Davies, A., McAteer, J., & Michie, S. (2010). Beliefs underlying UK parents’ views toward MMR promotion interventions: A qualitative study. Psychology, Health & Medicine, 15(2), 220–230.

Goldspink, C. (2010). Normative behaviour in Wikipedia. Information, Communication & Society, 13(5), 652–673.

Hallgren, K. A. (2012). Computing inter-rater reliability for observational data: An overview and tutorial. Tutor Quantitative Methods Psychology, 8(1), 23–34. Retrieved from http://www.ncbi. nlm.nih.gov/pmc/articles/PMC3402032/pdf/nihms372951.pdf

Hara, N., & Doney, J. (2015). Social construction of knowledge in Wikipedia. First Monday, 20(6). Retrieved from http://firstmonday.org/ojs/index.php/fm/article/view/5869

Hara, N., & Sanfilippo, M. R. (in preparation). Analysis of roles in knowledge co-construction of a controversial topic in child vaccination.

Hara, N., Shachaf, P., & Hew, K. F. (2010). Cross-cultural analysis of the Wikipedia community. Journal of American Society for Information Science & Technology, 61(10), 2097–2108.

Hargadon, A. B., & Bechky, B. A. (2006). When collections of creative become creative collectives: A field study of problem solving at work. Organization Science, 17(4), 484–500.

Hew, K. F., & Hara, N. (2007). Knowledge sharing among professionals in three online environ- ments: A qualitative case study. Journal of American Society for Information Science & Technology, 58(14), 2310–2324.

Holton, A., Weberling, B., Clarke, C. E., & Smith, M. J. (2012). The blame frame: Media attribution of culpability about the MMR-Autism vaccination scare. Health Communication, 27(7), 690– 701.

Jang, S. M. (2014). Seeking congruency or incongruency online? Examining selective exposure to four controversial science issues. Science Communication, 36(2), 143–167.

John, S. (2011). Expert testimony and epistemological free-riding: The MMR controversy. The Philosophical Quarterly, 61(244), 496–517.

Joyce, E., Butler, B., & Pike, J. (2011). Handling flammable materials: Wikipedia biographies of living persons as contentious objects. Proceedings of the 2011 iConference, Seattle, WA, pp. 25–32.

Joyce, E., Pike, J., & Bulter, B. (2013). Keeping eyes on the prize: Officially sanctioned rule breaking in mass collaboration systems. Proceedings of the 2013 conference on computer supported coopera- tive work, San Antonio, TX, pp. 1081–1092.

Kane, G. C., Johnson, J., & Majchrzak, A. (2014). Emergent life cycle: The tension between knowl- edge change and knowledge retention in open online coproduction communities. Management Science, 60(12), 3026–3048. Retrieved from http://dx.doi.org/10.1287/mnsc.2013.1855

König, R. (2013). Between lay participation and elite knowledge representation. Information, Communication & Society, 16(2), 160–177.

Kuk, G. (2006). Strategic interaction and knowledge sharing in the KDE developer mailing list. Management Science, 52(7), 1031–1042.

Lampe, C., Vitak, J., Gray, R., & Ellison, N. (2012). Perceptions of facebook's value as an information source. Proceedings of the 2012 annual conference on human factors in computing systems, Austin, TX.

Larson, H. J., Smith, D. M. D., Paterson, P., Cumming, M., Eckersberger, E., Freifield, C. C., … Madoff, L. C. (2013). Measuring vaccine confidence: Analysis of data obtained by a media sur- veillance system used to analyse public concerns about vaccines. The Lancet, 13, 606–613.

16 N. HARA AND M. R. SANFILIPPO

D ow

nl oa

de d

by [

N ew

Y or

k U

ni ve

rs it

y] a

t 06

:4 2

09 F

eb ru

ar y

20 16

Liao, Q. V., Wagner, C., Pirolli, P., & Fu, W. (2012). Understanding experts’ and novices’ expertise judgment of twitter users. Proceedings of the 2012 annual conference on human factors in com- puting systems, Austin, TX.

Luo, X. (2002). Trust production and privacy concerns on the Internet: A framework based on relationship marketing and social exchange theory. Industrial Marketing Management, 31(1), 111–118.

Majchrzak, A., Cooper, L. P., & Neece, O. E. (2004). Knowledge reuse for innovation. Management Science, 50(2), 174–188.

Majchrzak, A., Wagner, C., & Yates, D. (2013). The impact of shaping on knowledge reuse for organizational improvement with wikis. MIS Quarterly, 37(2), 455–469.

Markus, M. L. (2001). Toward a theory of knowledge ruse: Types of knowledge reuse situations and factors in reuse success. Journal of Management Information Systems, 18(1), 57–93.

Meservy, T. O., Jensen, M. L., & Fadel, K. J. (2014). Competing candidate solutions in electronic networks of practice. Information Systems Research, 25(1), 15–34.

Nagar, Y. (2012). What do you think? The structuring of an online community as a collective-sense- making process. Proceedings of the ACM 2012 conference on computer supported cooperative work 2012, Seattle, WA, pp. 393–402.

Nov, O. (2007). What motivates Wikipedians? Communications of the ACM, 50(11), 60–64. Petts, J., & Niemeyer, S. (2004). Health risk communication and amplification: Learning form the

MMR vaccination controversy. Health, Risk & Society, 6(1), 7–23. Promisel, A. G., & Centinaro, N. (2014). Company overview. Baby Center. Retrieved from http://

www.babycenter.com/help-about-company Purssell, E. (2004). Exploring the evidence surrounding the debate on MMR and autism. British

Journal of Nursing, 13(14), 834–838. Qualman, E. (2012). Socialnomics: How social media transforms the way we live and do business.

Hoboken, NJ: Wiley. Rechavi, A., & Rafaeli, S. (2012). Knowledge and social networks in Yahoo Answers. Proceedings of

the 45th HICSS annual conference, Maui, HI, pp. 781–790. Schroeder, A., & Wagner, C. (2012). Governance of open content creation: A conceptualization and

analysis of control and guiding mechanisms in the open content domain. Journal of the American Society for Information Science & Technology, 63(10), 1947–1959.

Seebach, C. (2012). Searching for answers ‒ Knowledge exchange through social media in organiz- ations. Proceedings of the 45th Hawaii international conference on system sciences, Maui, HI, pp. 3908–3917.

Shachaf, P., & Hara, N. (2010). Beyond vandalism: Wikipedia trolls. Journal of Information Science, 36(3), 357–370.

Shelby, A., & Ernst, K. (2013). Story and science: How providers and parents can utilize storytelling to combat anti-vaccine misinformation. Human Vaccines & Immunotherapeutics, 9(8), 1795– 1801.

Skea, Z. C., Entwistle, V. A., Watt, I., & Russell, E. (2008). ‘Avoiding harm to others’ considerations in relation to parental measles, mumps, and rubella (MMR) vaccination discussions: An analysis of an online chat forum. Social Science & Medicine, 67, 1382–1390.

Swarts, J. (2009). The collaborative construction of ‘fact’ on Wikipedia. Proceedings of the 27th ACM international conference on design of communication (SIGDOC’09), Bloomington, IN.

Tapscott, D., & Williams, A. D. (2006). Wikinomics: How mass collaboration changes everything. New York, NY: Portfolio.

Wiggins, A., & Crowston, K. (2011). From conversation to crowdsourcing: A typology of citizen science. Proceedings of the 44th Hawaii international conference on system sciences, Kauai, HI, pp. 1–10.

Wyatt, S., Bier, J., Harris, A., & van Heur, B. (2013). Participatory knowledge production 2.0: Critical views and experiences. Information, Communication & Society, 16(2), 153–159.

Yates, D., Wagner, C., & Majchrzak, A. (2010). Factors affecting shapers of organizational wikis. Journal of the American Society for Information Science and Technology, 61(3), 543–554.

INFORMATION, COMMUNICATION & SOCIETY 17

D ow

nl oa

de d

by [

N ew

Y or

k U

ni ve

rs it

y] a

t 06

:4 2

09 F

eb ru

ar y

20 16

Appendix: Inter-rater reliability Cohen’s κ

Content

Code Baby Center Berkeley Parents Network Wikipedia

Editorial 1 1 0.88 Background knowledge 0.43 0.52 0.66 Definitions 1 1 1 Facts about the topic 0.83 0.85 0.64 Media report about the topic 0.84 0.71 0.90 Resources 1 0.63 0.49 Citations 1 0.66 0.80 Claimed expertise 1 0.78 1 Confusion 1 0.46 0.85 Credibility 0.84 1 0.70 Experience 0.94 1 1 Inquiry 0.62 0.90 0.47 Interpretation or opinion 0.79 0.64 0.87 Vandalism 1 1 1

Role

Code Baby Center Berkeley Parents Network Wikipedia

Distractor 0.88 1 0.56 Giver 0.51 0.63 0.56 Seeker 0.79 1 0.64 Mover 0.49 0.57 0.57 Judge 0.86 1 0.61 Knowledge shaper 1 1 1 Organizer 0.49 1 1 Reframer 0.48 0.70 0.66 Reinforcer 0.56 0.63 0.5 Connector 0.66 1 0.79 Facilitator 0.66 1 1 Governor 1 1 0.55 Mediator 0.51 1 0.65 Supporter 0.56 0.63 0.56 Unmasker 0.88 1 1

18 N. HARA AND M. R. SANFILIPPO

D ow

nl oa

de d

by [

N ew

Y or

k U

ni ve

rs it

y] a

t 06

:4 2

09 F

eb ru

ar y

20 16