University of Kentucky UKnowledge
Theses and Dissertations--Communication Communication
2016
Social Media And Credibility Indicator: The Effects Of Bandwagon And Identity Cues Within Online Health And Risk Contexts
Xialing Lin University of Kentucky, [email protected] Digital Object Identifier: http://dx.doi.org/10.13023/ETD.2016.059
Right click to open a feedback form in a new tab to let us know how this document benefits ou.y
Recommended Citation Lin, Xialing, "Social Media And Credibility Indicator: The Effects Of Bandwagon And Identity Cues Within Online Health And Risk Contexts" (2016). Theses and Dissertations--Communication. 46. https://uknowledge.uky.edu/comm_etds/46
This Doctoral Dissertation is brought to you for free and open access by the Communication at UKnowledge. It has been accepted for inclusion in Theses and Dissertations--Communication by an authorized administrator of UKnowledge. For more information, please contact [email protected]. STUDENT AGREEMENT:
I represent that my thesis or dissertation and abstract are my original work. Proper attribution has been given to all outside sources. I understand that I am solely responsible for obtaining any needed copyright permissions. I have obtained needed written permission statement(s) from the owner(s) of each third-party copyrighted matter to be included in my work, allowing electronic distribution (if such use is not permitted by the fair use doctrine) which will be submitted to UKnowledge as Additional File.
I hereby grant to The University of Kentucky and its agents the irrevocable, non-exclusive, and royalty-free license to archive and make accessible my work in whole or in part in all forms of media, now or hereafter known. I agree that the document mentioned above may be made available immediately for worldwide access unless an embargo applies.
I retain all other ownership rights to the copyright of my work. I also retain the right to use in future works (such as articles or books) all or part of my work. I understand that I am free to register the copyright to my work.
REVIEW, APPROVAL AND ACCEPTANCE
The document mentioned above has been reviewed and accepted by the student’s advisor, on behalf of the advisory committee, and by the Director of Graduate Studies (DGS), on behalf of the program; we verify that this is the final, approved version of the student’s thesis including all changes required by the advisory committee. The undersigned agree to abide by the statements above.
Xialing Lin, Student
Dr. Patric Spence, Major Professor
Dr. Bobi Ivanov, Director of Graduate Studies SOCIAL MEDIA AND CREDIBILITY INDICATOR: THE EFFECTS OF BANDWAGON AND IDENTITY CUES WITHIN ONLINE HEALTH AND RISK CONTEXTS
DISSERTATION
A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the College of Communication and Information at the University of Kentucky
By Xialing Lin
Lexington, Kentucky
Director: Dr. Patric Spence, Associate Professor, School of Information Science
Lexington, Kentucky
2016
Copyright © Xialing Lin 2016 ABSTRACT OF DISSERTATION
SOCIAL MEDIA AND CREDIBILITY INDICATOR: THE EFFECTS OF BANDWAGON AND IDENTITY CUES WITHIN ONLINE HEALTH AND RISK CONTEXTS
Three studies were conducted to investigate how social media affordances influence individuals’ source credibility perceptions in risk situations. The MAIN model (Sundar, 2008), warranting theory (Walther & Parks, 2002), and signaling theory (Donath, 1999) served as the theoretical framework to examine the effects of bandwagon cues and identity cues embedded in retweets and users’ profile pages for health and risk online information processing. Study One examines whether bandwagon heuristics triggered by retweets would influence individuals’ source credibility judgments. Study Two investigates how bandwagon heuristics interact with different identity heuristics in credibility heuristics on an individual level. Study Three explores bandwagon heuristics at the organizational level. Three post-test only experiments with self-report online surveys were conducted to investigate the hypothesis and research questions. Results indicate that different online heuristic cues impact the judgments of competence, goodwill, and trustworthiness at different levels. Authority strongly influenced source credibility perceptions. A reverse-bandwagon effect was observed in influencing source credibility judgments. Theoretical and practical implications are discussed.
KEYWORDS: MAIN Model, Bandwagon Heuristic, Identity Heuristics, Risk and Health, Twitter
Xialing Lin
April 19, 2016 Date SOCIAL MEDIA AND CREDIBILITY INDICATOR: THE EFFECTS OF BANDWAGON AND IDENTITY CUES WITHIN ONLINE HEALTH AND RISK CONTEXTS
By
Xialing Lin
Dr. Patric Spence Director of Dissertation
Dr. Bobi Ivanov Director of Graduate Studies
April 19, 2016
ACKNOWLEDGMENTS
Completion of this dissertation was impossible without the insights and support from several people. First, my advisor and dissertation chair, Dr. Patric Spence, plays a fundamental role in my professional and personal development. My Dissertation Committee and outside examiner, Dr. Timothy Sellnow, Dr. Deanna Sellnow, Dr. Xin Ma, and Dr. April Young, exemplify high-quality expertise that keeps inspiring me. I also wish to thank my family and friends, who always come with full of optimism and encouragement, providing me a positive social environment. Each of these individuals has demonstrated wisdom and goodness that guided and challenged my thinking, substantially lead me towards the completion of the doctoral program.
Patric has been both of my master and doctoral advisor since 2011; I have been working with him for more than five years. He has been a tremendous mentor for me in both research, teaching, and life experience. Patric accepted me as his advisee when I was new to academia and struggling with acculturative stress. He provides me with every bit of guidance, assistance, proficiency, and patience, which encouraged me to build up my competence and confidence for my career. He has been involving me in his projects and generously bridges me to his networking; from there I have opportunities to construct academic collaborations with him and other outstanding scholars, such as Dr. Ken Lachlan, Dr. Tim Sellnow, and Dr. David Westerman.
Patric gave me the freedom to venture into research, at the same time continuingly contributes valuable feedback, advice, and encouragement. He helped me to narrow down my dissertation foci as early as possible, set achievable research questions and objectives when I felt ready to venture into my dissertation, and provide academic and life advice whenever needed. He inspires me to grow into a risk and crisis research scientist. In addition to our academic collaboration, I greatly value the personal rapport that Patric and I have forged over the years.
I am also very thankful to my committee members, Tim, Deanna, and Dr. Ma, for their time and valuable feedback on a preliminary version of this work. Tim and Deanna provided thoughtful suggestions and directions as early as the time I was preparing my doctoral examination, which contributes significant insights into the drafting of the literature review and discussion chapters. I am also indebted to Dr. Ma, who deliberately advised on my methods. It was a joyful experience to having all of them in my Doctoral Committee.
It is fortunate for me to have Tim’s risk communication and Dr. Ma’s research method classes, which have had a substantial influence on my research. I also would like to acknowledge Deanna, who has been an inspiring role model for me. She and Tim show great wisdom, patience, and cheering, which also encourage me to undergird all of my endeavors.
I also owe my appreciation to my parents. Although they are not physically by my side, they unconditionally support me at every stage of my personal life and longed to see
iii
this achievement come true. They respect every of my decision, gave me freedom to branch out into new life journey when I was little and pursuit whom I want to be. Words cannot express how grateful I am to my mother and father for all of the sacrifices they have made on my behalf. Their prayer for me has sustained me thus far.
Last but not least, some faculty members, scholars, and friends have been kind enough to extend their help at various phases of my life whenever I approached them, and I do hereby acknowledge all of them. I thank Dr. Ken Lachlan, Dr. Bobi Ivanov, Dr. Amy Gaffney, and Dr. Brandi Frisby for all their goodness.
iv
TABLE OF CONTENTS CHAPTER 1: INTRODUCTION ...... 1 Social Media and Crisis/Risk Communication ...... 1 Theoretical Framework Overview ...... 3 Organization of the Dissertation ...... 7 CHAPTER 2: LITERATURE REVIEW ...... 9 The Roles of Social Media during Risk and Crises ...... 10 Source Credibility and Its Role in Risk and Crises ...... 13 Twitter and Heuristic Processing ...... 15 Warranting Theory ...... 20 Signaling Theory...... 24 The MAIN Model ...... 26 Bandwagon Heuristics and the Numbers of Retweets ...... 29 Identity Heuristic, and Interactions with Bandwagon ...... 31 Source Credibility Dimensions and Measurement ...... 34 Chapter Summary ...... 37 CHAPTER 3: METHODS ...... 38 Study One ...... 39 Participants ...... 39 Materials ...... 39 Measures ...... 40 Study Two ...... 41 Participants ...... 41 Materials ...... 42 Measures ...... 43 Study Three ...... 44 Participants ...... 44 Materials ...... 45 Measures ...... 45 Chapter Summary ...... 47 CHAPTER 4: RESULTS ...... 48 Study One ...... 48
v
Sample Profile ...... 48 Confirmatory Factor Analysis on Source Credibility Scales ...... 48 Bandwagon Heuristics and Source Credibility ...... 49 Path Analyses ...... 50 Demographics and Twitter Usage ...... 51 Study Two ...... 53 Sample Profile ...... 53 Confirmatory Factor Analysis on Source Credibility Scales ...... 53 Source Credibility Perceptions ...... 54 Study Three ...... 57 Sample Profile ...... 57 Source Credibility Perceptions on Organizational Level ...... 57 Chapter Summary ...... 58 CHAPTER 5: DISCUSSIONS ...... 65 General Discussion ...... 65 Study One ...... 66 Primary Findings ...... 66 Limitations and Future Directions ...... 68 Study Two ...... 70 Theoretical Implications ...... 71 Practical Implications ...... 76 Limitation and Future Directions ...... 78 Study Three ...... 79 Theoretical Implications ...... 80 Practical Implications ...... 83 Limitations and Future Directions ...... 85 Chapter Summary ...... 87 CHAPTER 6: CONCLUSION ...... 88 APPENDICES ...... 90 Appendix A: Manipulation Example for Study One ...... 90 Appendix B: Authority Manipulation Example for Study Two ...... 91 Appendix C: Peer Manipulation Example for Study Two ...... 92
vi
Appendix D: Stranger Manipulation Example for Study Two ...... 93 Appendix E: Manipulation Example for Study Three ...... 94 References ...... 95 CURRICULUM VITAE ...... 111
vii
LIST OF TABLES Table 4.1, Study One: Demographic Characteristics of Participants ...... 59 Table 4.2, Study Two: Demographic Characteristics of Participants ...... 60 Table 4.3, Study Three: Demographic Characteristics of Participants...... 61 Table 4.4.1, Study Two: Descriptive Statistics for Competence as a Function of User Identities and Retweet Numbers ...... 62 Table 4.4.2, Study Two: Tukey Post-hoc Test Multiple Comparisons for User Identity on Competence ...... 62 Table 4.4.3, Study Two: Tukey Post-hoc Test Multiple Comparisons for Retweet Numbers on Competence ...... 62 Table 4.5.1, Study Two: Descriptive Statistics for Trustworthiness as a Function of User Identities and Retweet Numbers ...... 63 Table 4.5.2, Study Two: Tukey Post-hoc Multiple Comparisons for User Identity on Trustworthiness ...... 63 Table 4.5.3, Study Two: Tukey Post-hoc Multiple Comparisons for Retweet Numbers on Trustworthiness ...... 63 Table 4.6.1, Study Two: Descriptive Statistics for Goodwill as a Function of User Identities and Retweet Numbers ...... 64 Table 4.6.2, Study Two: Tukey Post-hoc Test Multiple Comparisons for User Identity on Goodwill ...... 64 Table 4.7, ANOVA and Tukey Post-hoc Test Results among Three Conditions ...... 64
viii
CHAPTER 1: INTRODUCTION
Social Media and Crisis/Risk Communication
Social media have been transforming the ways that individuals process
information during risk and crisis situations. Unlike traditional news media which often
depend on professional gatekeepers to disseminate information, social networking sites
provide multi-source environments, which allow individuals to perform as gate watchers
to co-create user-generated content (Metzger, Flanagin, & Medders, 2010). First reports from eyewitnesses often use social media outlets as an efficient avenue when reporting events as they are taking place (Spence, Lachlan, Lin, & Del Greco, 2015). Previous empirical cases have shown that even traditional news media outlets glean information from social media as “backchannel news” during crisis situations (Spence et al., 2015a).
For example, during the 2010 Haitian earthquake, social media played a key role in
disseminating information in the absence of traditional media outlets that had been
compromised by power outages (Bunz, 2010). As one of the mainstream social
networking and microblogging services, Twitter has gained increased recognition as a legitimate information source and has been playing a significant role during crises and other extreme events (Lachlan, Spence, & Lin, 2014).
The technology affordances on Twitter make immediate access and real-time eyewitness accounts possible in equivocal and highly unpredictable conditions (Lachlan
et al., 2014b; 2014c; Spence et al., 2015a). By initially providing users the opportunity to post, read, and respond to text-based messages limited to 140-characters in length,
Twitter creates a multi-media platform with constantly updated timelines for wide-open
content. These 140-character messages, called tweets, range from life chores to breaking
1
news. Twitter has accumulated around 302 million monthly active users as of the writing
of this manuscript, who connect with each other by following or being followed without reciprocal requirements (twitter.com). Users can follow to view any other’s information in their Twitter streams without bilateral consent, while also having their own groups of followers. As the platform grew, some specific features evolved for tweets: the “RT” which stands for retweet, means to repost a message from another Twitter user and share
it with one's own followers; users could contain the other’s username in a tweet preceded
by the “@” symbol to mention other users, as well as reply to another users’ tweets with
the “@” symbol followed by the recipient's username.
The symbol of “#” prefixing keywords or phrases in a tweet is known as a
hashtag, and allows users to categorize posts together by topics or types, and provide
links for easy Twitter searching. Twitter has continuously modified its interface for a
more user-friendly and news-updated design. Twitter can be used for numerous
applications for mobile phones and tablets, making it a useful medium for
communicating during crises when other means of communicating may be unavailable.
These media technological functions are embedded with sets of underlying affordances
that can shape the nature and the presence of content, which in turn influence people’s
credibility processing online (Sundar, 2008; Csikszentmihalyi & Kubey, 1981).
The public has gradually used Twitter as a source throughout the crisis lifecycle
to make sense of the event and to make well-informed decisions, especially for natural
disasters that cause varying degrees of destruction or harm to vulnerable populations. For
instance, Spence et al., (2015a) analyzed the content and the frequency of over 1,500
tweets being sent during the prodromal state of 2012 Hurricane Sandy, and found that
2
Twitter served as a medium for information sharing and an outlet for affective
expressions (Spence et al., 2015a). As the identity of information sources online “is
usually murky” especially in high uncertainty situations, it is critical for online users to
evaluate the credibility of information sources (Sundar, 2008, p. 83). Compared to the
proliferation of Twitter during disasters and other extreme events, a major research
question revolving around how people assess source credibility on Twitter still remains for further attention (Westerman, Spence, & Van der Heide, 2014). In particular, the investigation of how different online heuristic cues affect source credibility perceptions
in risk and crisis situations is still in its infancy (Van Der Heide & Lim, 2015). Thus, the
current studies intend to investigate how Twitter affordances influence individuals’
source credibility perceptions in risk situations. More specifically, the effects of
bandwagon cues and identity cues, embedded in retweets and users’ profile pages, were
investigated for risk information source credibility assessments.
Theoretical Framework Overview
The MAIN model (Sundar, 2008), warranting theory (Walther & Parks, 2002),
and signaling theory (Donath, 1999) are used as the theoretical framework for the current
study. The MAIN model (Sundar, 2008) outlines how four technological affordances,
including modality, agency, interactivity, and navigability that are presented in most
social media, influence individuals’ online credibility assessments. Given information
overload and cue substitutability online, individuals tend to engage in heuristic
processing as it requires the least effort; thus, effortless information judgments are
triggered prior to more systematic, cognitive analysis (Chen & Chaiken, 1999; Walther,
1992; Metzger et al., 2010). To reduce risk-related uncertainties during extreme events,
3
online users would apply information sources or design features on media platforms not
only to invoke heuristic processing, but also to assist subsequent systematic information seeking (Sundar, 2008). According to the MAIN model, the agency affordances in the interface of Twitter could trigger various heuristic processing, such as identity heuristics or machine heuristics, to guide the public toward accurate credibility judgments (Sundar,
2008). In this study, the effects of identity and machine heuristics, related to agency
affordances are investigated.
Identity cues are those that trigger credibility perceptions based on the allowance
of users’ self-identity assertion through agency affordances on social media (Sundar,
2008). Given information anonymity and aggregated authorship online, information
associated with individual identities, such as tweets posted by a peer or an organization,
may assist online users to judge the message content quality (Walther, Van Der Heide,
Tong, Carr, & Atkin, 2010; Sundar, 2008). Machine heuristics are attributed to the
system generated cues which imply randomness, objectivity, and other media mechanical
characteristics (Westerman et al., 2014; Sundar, 2008). Previous studies suggest that
individuals consider machine heuristics as robust cognitive shortcuts and result in
positive credibility assessments (Sundar, 2008). In addition, machine heuristics may
serve to cue other heuristics influenced with user identities or the large user base. For
instance, the retweet feature on Twitter is displayed as numbers under the original tweet
on a user’s profile page. The numbers of retweets would work as a machine cues which
also invoke bandwagon heuristics. The notion of machine heuristics are consistent with
warranting theory (Walther & Parks, 2002) and signaling theory (Donath, 1999).
4
Warranting theory (Walther & Parks, 2002) indicates an interactive ecology
online and suggests that communication affordances allow individuals to create selective
or deceptive self-presentations. The less extent the information can be manipulated by message senders, the higher levels of validity or authenticity the message would have.
According to warranting theory, information presented by a sender’s network is
considered to be more credible than information presented by the sender. Therefore, the
machine-generated heuristic cues may serve as warranting values which validate the
content of the messages online. Information endorsed by retweets may be more likely to
be seen as trustful to online users to the extent that the online users may perceive it to be
immune to manipulation by the owner of the original tweets (Walther, 2011; Walther &
Parks, 2002). This theory helps to explain the dynamics of the assessment on the
veridicality of online interactions and individuals’ self-presentations.
Consistently, Donath (1999) suggests a theoretical basis underlying the skepticism
about the legitimacy of online users’ self-presentation. According to signaling theory
(Donath, 1999, 2007), when a signal or cue cost much to construct or produce, online
users would consider the signal or information cue as reliable; the interactions with other
users in a target’s social network would increase source credibility perception of the
target (Walther, 2011). The notion of signaling theory (Donath, 1999, 2007) could also be
used to explain the effects of machine heuristics such as retweets on judgments of source
credibility.
Although previous research provides significant insights into the function of
retweets on source credibility assessment, limited numbers of studies investigate this
within risk and crisis contexts. Based on the above-mentioned theoretical constructs, the
5 current set of studies focuses on the underlying affordances rather than superficial interface functions on certain social media. More specifically, these studies focus on risk and crisis situations, to examine whether bandwagon heuristics triggered by retweets would influence individuals’ source credibility judgments, and the extent to which retweet numbers would affect individuals’ judgments. Moreover, these studies investigate how bandwagon heuristics interact with different identity heuristics and credibility heuristics on an individual level; this study also investigates whether the bandwagon heuristics emerge at the organizational level.
Three experiments were conducted to investigate the research questions. A fictional food safety topic within health risk forum was used as the risk and crisis manipulation in the experiments. The first experiment intends to investigate whether retweets affect source credibility assessments on Twitter. A two-condition design with retweets (present vs. absent) on a mock stranger’s Twitter profile page was created. The second experiment intends to investigate the extent to which the number of retweets in influencing source credibility perceptions and its interactions with different individuals’ identity cues on Twitter. A 3 (the numbers of retweets of an original tweet: 40 vs. 400 vs.
4,000) × 3 (users’ identity cues: stranger’s vs. peer’s vs. authority’s individual profile pages) were created. The third experiment intends to examine whether bandwagon heuristics invoked by retweets would also influence Twitter users’ credibility assessments when the original tweets were posted by an organizational account. A three-condition design with the numbers of retweets (i.e., 40 vs. 400 vs. 4,000) was presented under an original tweet on an organizational account’s profile page.
6
To ensure the measurement validity and reliability in the study, McCroskey and
Teven’s (1999) Source Credibility scales with three separate constructs (i.e., competence, goodwill, and trustworthiness) were applied to measure source credibility in the first two experimental designs, whereas the RAND Public Health Disaster trust scale (Eisenman et al., 2012) was utilized in the third experiment. The McCroskey and Teven (1999) scales were rooted from interpersonal communication and designed to assess dispositions toward identifiable individuals (O’Keefe, 1990). These scales contain three separate constructs, including competence, goodwill, and trustworthiness (O’Keefe, 1990).
Previous studies suggest that trustworthiness and institutional trust are different concepts, as individual credibility is more related to firsthand knowledge while organizational credibility is more associated with a general reputation (Eisenman et al., 2012). Thus, the
RAND Public Health Disaster Trust scale was used to facilitate identifying credibility in organizations within the risk and crisis scenario on Twitter for the third study.
Organization of the Dissertation
This dissertation consists of five chapters. Research background, problems, and designs was outlined in Chapter 1. In Chapter 2, a comprehensive literature review in light of the MAIN model, warranty theory, and signaling theory was provided; the research significance and the research questions were elaborated. Chapter 3 explained the research methods, including conceptual operationalization, participant recruitments, experiment designs, the implementation of the experiments, and measurement instruments. Chapter 4 analyzed the data and reported the results. The research discussions regarding the study results, theoretical and practical implications, as well as research limitation and future directions, were provided in Chapter 5. A brief conclusion
7 is provided in Chapter 6. In the proceeding sections, literature relating to the MAIN model, warranting theory, signaling theory and credibility in a risk crisis context was reviewed. Three research questions relating to the areas of reviewed literature were offered, in addition to the proposed experimental procedures.
Copyright © Xialing Lin 2016
8
CHAPTER 2: LITERATURE REVIEW
Crises, risks, and other extreme events are unfolding processes leading to increased uncertainties and perceived threats, which typically require individuals, stakeholders, and communities “to make the best possible decisions about their well- being” with limited information during tight time constraints (Reynolds, 2006, p.6).
Many empirical lessons, such as the 2012 “pink slime” labels for the use of Lean Finely
Texture Beef (LFTB), have highlighted the need to establish effective risk and crisis communication with the public (Spence, Lachlan, Lin, Sellnow-Richmond, & Sellnow,
2015). Such communication needs to be accurate, timely, interactive, and collaborative
(Seeger, 2006; Sellnow, Ulmer, Seeger, & Littlefield, 2009; Spence et al., 2015b). With the prevalence of social media, the contemporary information landscape has changed over the past decade regarding information transmission and assessment, which in turn affects the application of risk reduction and crisis management.
The amalgamation of social media and emergency communication has experienced increased research attention (Freberg & Palenchar, 2013). Despite the advances empowered by innovative social media technologies, the use of new media platforms also alter offline information processing and create challenges for the public in locating credible and trustworthy information, especially during risk and crisis situations.
As social media has gradually been playing a more critical role in risk and crisis management, it becomes more and more important to understand how people process online risk information for both the public and health/emergency practitioners. Guided by literature in communication technology and risk and crisis research, this chapter reviews source credibility assessment within social media affordances with a specific
9 focus on Twitter. This focus includes: (1) the roles of social media during risk and crises,
(2) online source credibility perceptions, (3) Twitter and heuristic information processing,
(4) warranting theory, signaling theory, and the MAIN model as theoretical frameworks, and (5) source credibility dimensions and measurement. Research questions and hypotheses are also forwarded.
The Roles of Social Media during Risk and Crises
Social media, such as Twitter, Facebook, and short message services (SMS) are a group of interactive, collaborative, conversational, and community-based systems built upon “the ideological and technological foundations of Web 2.0” (Kaplan & Haenlein,
2010, p. 61; Mayfield, 2006). Social media collapses diverse social contexts and media audiences into one, providing a comprehensive information outlet not only for daily life, but also for threat situations with high uncertainty (Westerman et al., 2014; Lachlan,
Spence, Lin, Del Greco, & Najarian, 2014). A survey conducted by Pew reported that over 73% of adult Americans were online information users, 80% of whom have sought health information online (Fox, May 12, 2011). Public health practitioners have been applying social media to promote public health-related issues, including sexually transmitted diseases (STDs) prevention among homosexual partners, condom use, HIV prevention, and Chlamydia prevention (Jones, Baldwin, Lewis, 2012; Purdy, 2011; Rice
Tulbert, Cederbaum, Adhikari, Milburn, 2012). For instance, in recent years, the Centers for Disease Control and Prevention (CDC) and nongovernmental organizations have also paid increased attention to the use of social media, building guidance protocols for epidemic tracking and health message dissemination (e.g., CDC Social Media Tools,
Guidelines and Best Practices, 2014, January 22). Compared to traditional or legacy
10
media, social media have revolutionized the way people communicate and process
information within many contexts, which could either facilitate or thwart the public’s risk and crisis resilience (Spence et al., 2015a; Walther et al., 2010).
Holding great promise as a news channel for crisis and risk communication, social media provide publicly available online information, reflecting immediate accessibility and harnessing collective intelligence (O’Reilly & Battelle, 2009; Wunsch-Vincent &
Vickery, 2006). These new media platforms serve as a legitimate outlet from diverse sources, and back-up overwhelmed traditional media that fail to cope with the heavy traffic in disseminating the latest information during emergencies (Lachlan et al., 2014c;
Pepitone, 2010). For instance, Facebook and other social media were used as a dominant source of news outlets to share warning messages during the 2010 volcano eruptions in
Iceland and the 2011 Queensland floods in Australia (Reuter, 2015; Wendling, Radisch,
& Jacobzone, 2013). Social media allow citizens to communicate directly their experiences to general public audiences during the unfolding of crises, which generates community crisis maps and empowers community resilience (Goolsby, 2010). Local residents often act as eyewitnesses to report these events on social media in a timely manner, such as the 2010 Haiti cholera outbreaks, the 2011 tsunami disaster in Japan, and Hurricane Sandy, which occurred in 2012 (e.g., Dredze, 2012; Lachlan et al., 2014c;
Sutter, 2010). Considering the advances of social media, risk and crisis experts, practitioners and official agencies have gradually adapted these new media technologies as a main source for monitoring local threats, communicating emergency alerts, providing official crisis response guidance in the form of instructional content,
11
communicating with stakeholders, and appealing for assistance (Lachlan, Spence, Lin, &
del Greco, 2015).
Despite the increasing utilization of social media during natural disasters, new
media ecology also represents new challenges for the public to locate information that
they believe is credible and that they can trust. Although previous studies revealed a
higher level of perceived credibility in active updates on social media rather than in
traditional media crisis coverage, the assurance of online content quality is still
questionable for many (Horrigan & Morris, 2005; Procopio & Procopio, 2007; Sundar,
2008). Online content is user-generated, originating from various venues that are heterogeneous in the quality, origin, and veracity of information (Sundar, 2008; Walther
et al., 2010). The vast repository and source fecundity of online information promote
concerns about its credibility and accuracy, which would, in turn, affect people
discerning and interpreting this information (Lachlan et al., 2015; Metzger & Flanagin,
2015).
In many risk and crisis situations where knowledge is informal, uncertain, and
diffused by word-of-mouth, it is often the case that the most timely and reliable
information is gleaned not from a traditional source, but rather from a diversity of lay
people lacking specific training, credentials, or established reputations (Pure et al., 2012;
Spence, Lachlan, Burke, & Seeger, 2007). The new media ecology shifts the gatekeeping
function from professional content producers, such as editors imbued with authority, to
content consumers, which directly transfers the task of assessing information accuracy
and credibility to online users (Westerman et al., 2012). Such a transition has shifted the
burden of factual verification and content analysis onto media users. Previous studies
12
indicate that people are compelled to consume any updated news they could obtain
during extreme situations. People are increasingly relying on social media that contain a
vast diversity of sources during risk and disasters, even when the disseminated
information were unconfirmed or inaccurate (Freberg & Palenchar, 2013; Spence et al.,
2015a; Westerman et al., 2014). However, the unregulated flow of information online is
problematic for those who seek information, as it becomes harder to distinguish more
credible sources from less credible ones (Eastin, 2006). Given that the high stakes
associated with misinformation during risk and disasters have potentially threatening consequences, it is critical to provide evidence-based guidelines for source credibility judgment on social media.
Source Credibility and Its Role in Risk and Crises
The research on source credibility has a long history, and the research has been both interdisciplinary and pointed out the construct as a joined concept. In psychology and communication literature, credibility refers to the believability of information, which treats credibility as a receiver-based construct and emphasizes a subjective judgment and perception on the part of a receiver concerning accuracy, reliability, currency and comprehensiveness of the information (Fogg & Tseng, 1999; Rieh & Danielson, 2007).
Some research has been investigating various foundations from which credibility can arise and focus attention on the characteristics of the source (Metzger, Flanagin, Eyal,
Lemus, & Mccann, 2003). Source credibility is defined as “judgments made by a perceiver concerning the believability of a communicator”, which is the most widely studied area for credibility (O’Keefe, 1990. p.130-131). Message credibility examines message characteristics, such as its structure, content, and language intensity, impact
13
perceptions of believability (Metzger et al., 2003). Moreover, media credibility is concerned about technology and structural features of a medium channel in influencing
perceptions of believability (Metzger et al., 2003; O’Keefe, 1990). All of these
conceptualizations focus on judgments, characteristics, content, and attitudes. Credibility
research has placed emphasis on the characteristics of the source as the concept of
credibility that rests largely on “the trustworthiness and expertise of the source or
messages” (Metzger & Flanagin, 2015, p. 446). Thus, this set of studies focuses on
source credibility and explores how online users perceive credibility from different
sources that are empowered by certain social media affordances during risk situations.
Thus, starting with the belief that the public has the same goals online that they would
have in face-to-face communication during a crisis, a central focus of this set of studies is
on the relationship between media affordances and credibility.
Source credibility is foundational to risk and crisis communication practices as it
has long been recognized as a key variable to message acceptance and interpretation
(Hovland, Janis, & Kelley, 1953; Reynolds & Seeger, 2005). Renn and Levine (1991)
suggest that credibility issues should be investigated in the contexts of the arena where
they occur. As many risk debates and crisis situations are characterized by new evolving
power structures and substantial values, source credibility becomes a major medium of
power control and social influence in such contexts (Renn & Levine, 1991). Empirical
evidence suggests that credibility works as a persuasive attribute that may significantly
reinforce the perceived legitimacy of the information providers as well as public
approvals and trustworthiness; lack of trust or credibility can doom risk and crisis
communication efforts, which may in turn increase the potential harm and severity of the
14 risk or crisis (Glik, 2007; Renn & Levine, 1991). For instance, studies show that racial and ethnic minorities tend to mistrust health and risk messages from official agencies in the aftermath of the anthrax attacks of 2001 whereas adherence to government directives was essential to public health and national security (Braveman, Egerter, Cubbin, &
Marchi, 2004; Quinn, Thomas, & McAllister, 2005). This highlights how essential it is to understand the process of source credibility evaluation in order to reach the public and stakeholders during extreme events.
Source credibility is suggested to be distinct from media organizations, channels, or the content of the news itself (Metzger et al., 2003). However, several scholars, such as
Chaffee (1982), indicate that there may be a conflation for different sources, media, and message credibility; receivers do not consciously distinguish a message’s source from its channel through which it is received (Metzger et al., 2003). These notions were particularly reflected in new media environments, as the source credibility of online information could be attributed to the users, technologies, and structures, as well as the operators or sponsors of the medium; thus, the source, media, and message credibility may commingle (Sundar, 1998; Sundar & Nass, 2000, 2001). It is, therefore, possible that different online attributions to an online message may influence online users to perceive credibility differently. In the next section, one of the mainstream social media platforms,
Twitter, is applied to understand how different media affordances influence credibility perceptions.
Twitter and Heuristic Processing
Twitter is one of the social networking sites and microblogging services launched in March of 2006. This social medium then rapidly broke into mainstream during 2009,
15 and has accumulated around 302 million monthly active users with 500 million tweets being posted per day at the time of the writing of this manuscript (twitter.com). By initially providing users the opportunity to post, read, and respond to text-based messages capped at 140 characters, Twitter creates a multi-media platform with constantly updated timelines for wide-open content. These messages, called tweets, range from life chores to breaking news.
Twitter users connect with each other by following or being followed without reciprocal requirement. Such non-reciprocal relationships are implemented by multiple mechanisms on Twitter, such as retweets, liking, and hashtags, making information dissemination and retransmission possible on a large scale (Sutton et al., 2014). Taking the function of the retweet as an instance, Twitter users can repost a message from another user and share it with one’s own followers. A previous study indicated that, in
2010, approximately 11% of all tweets were retweets, and users tend to retweet messages containing hashtags and URLs (Suh, Hong, Pirolli, & Chi, 2010). Online users can view the numbers of retweets, likes, and replies of each tweet on its sender’s profile page.
Particularly, twitter users could elect to share and disseminate information on the basis of retweet and reply mechanisms that amplify original sources (Kwak, Lee, Park, & Moon,
2010; Sutton et al., 2014).
Twitter has been widely adopted as a conduit for emergency responses and risk communication by the public and official authorities during high consequence events, supporting efficient, timely, and targeted communication efforts. These efforts and affordances of Twitter are capturing the interests of scholars and practitioners (Spence,
Lachlan, Lin, & Del Greco, 2015). Research on Hurricane Sandy in 2012 found that,
16 during the prodromal stage (or precrisis stage) leading up to the landfall of this major weather crisis, Twitter transitions from being a medium of providing information to an outlet for affective responses and expressions of fear (Spence et al., 2015). However, the results also indicate the challenges when people use social media to locate tangible instructional information concerning health and property in large-scale crises (Spence et al., 2015). Research on the Canadian Red River Valley floods of 2009 also suggests that civilians tend to share tweets with authoritative news sources or information from higher levels (Starbird, Palen, Hughes, & Vieweg, 2010). In spite of the noisy and unstructured user-generated content on Twitter, the tweets and their associated technological functions indicate the user-based sentiment and behaviors, which reflects the dynamics of various media affordances and mechanisms (Cheong & Lee, 2011). These studies outline the use of Twitter by various publics during risk and disasters and highlight the need for credibility research within social media.
Although previous studies examined the behavioral patterns that the public exhibit during risk and crises (e.g., Bean et al., 2016; Spence et al., 2015a; Lachlan et al., 2014a;
2014b; 2014c; Sutton, League, Sellnow, & Sellnow, 2015), it is unclear how people process risk and crisis information by certain media affordances. The tweets and their associated information, such as the updated time, location, numbers of retweets and liking, as well as users’ characteristics, exhibit the characteristics of social media affordances (Cheong & Lee, 2011). A widely-accepted premise for online credibility perception in the literature posits that individuals are more dependent on cues and heuristics than they would be in face-to-face communications (Walther & Jang, 2012).
This premise is central for the arguments in this series of studies.
17
Additionally, in light of social media technological affordances and the threat of
information overload, individuals tend to engage in cognitive heuristic processing with
the least effort possible; thus effortless information judgments are triggered prior to more
systematic, cognitive analysis (Chen & Chaiken, 1999; Walther, 1992; Metzger et al.,
2010). In order to reduce the risk-related uncertainty, the heuristic shortcuts on social media, such as source or design features, not only invoke heuristic processing, but may
also assist subsequent systematic information seeking (Sundar, 2008).
Machine heuristic. Sundar and his colleagues (Sundar, Jia, Waddell, & Huang,
2015) indicates that various characteristics of perceived sources may affect different
levels of credibility judgments. Consistently, individuals tend to treat different attributed
sources as peripheral cues, which may lead to cognitive heuristics or guide more
systematic processing of the content (Sundar, 2008). Specifically, source agents operated
by media affordances would cue the machine heuristic (Sundar, 2008). This heuristic
serves as a cognitive shortcut through which individuals assign greater credence to
information that is verified and/or chosen by system entities (Edwards, Spence, Gentile,
Edwards, & Edwards, 2013; Sundar, 2008). For example, a web counter at the bottom of
a web page could invoke this heuristic. The web counter simply indicates the number of
people to visit a website. The counter does not have a political affiliation or other bias,
and as long as the person viewing the counter does not believe it is being manipulated, it
could provide a credible indication of the popularity of a website. The machine heuristic
is featured as randomness, objective, and other mechanical characteristics to its
performance (Sundar, 2008).
18
Media agents low on humanness would generate positive trustworthiness in the
information due to the machine heuristics and lead to presumptions of objectivity
(Sundar, 2008; Sundar & Nass, 2001) and many aspects of social media can be viewed as
being low on humanness. Machine heuristics may impact the way that individuals
process system-generated cues online (Westerman et al., 2013). Previous research on
recency and updates of social media suggest that recency of tweets impacts source
credibility through cognitive elaboration (Westerman et al., 2014). Machine heuristics
may occur when system-generated cues create a situation that online users need to
consider more about or elaborate on, and this consideration or elaboration would lead to
higher judgments of credibility (Sundar, 2008; Westernman et al., 2013). Consistently,
Edwards et al. (2013) experimentally examined whether or to what degree Klout scores
from a social media page can influence perceptions of credibility. The Klout score is an aggregation of 400 signals from seven social media networks computed based on a user’s ability to drive action in social networks. The results demonstrated that people perceived the mock Twitter page with a high Klout score higher in credibility than the identical mock Twitter page with a moderate or low Klout score. Thus, like the web page counter,
Klout scores provide a non-human manipulated cue that can be used to judge credibility.
Based on the aforementioned notions, Sundar (2008) suggests the application of heuristic cues to effectively guide the public toward accurate credibility judgments. In order to explicate the effects of underlying media affordances in guiding source credibility evaluations within a Twitter context during risk and crises, warranting theory
(Walther & Parks, 2002), signaling theory (Donath, 1999), and the MAIN model (Sundar,
2008) are utilized as the theoretical framework for the current study.
19
Warranting Theory
As online self-presentation requires little effort to fabricate through mediated
communication; online users raise the suspicions about the authenticity and reliability of
the message content on new media (Caspi & Gorsky, 2006; Donath, 1999). Individuals
with access to a computer and the internet can create and disseminate messages. Given
the lack of immediate physical presence online, people may avail for dissembling and
selective self-presentations, causing online viewers may have difficulties to detect
deceptions (Frankel & Sang, 1999). Stated differently, a person can use the affordances
of online media to present any image of themselves they wish. Accordingly, warranting
theory (Walther & Parks, 2002) outlines the construct of warranting value to explicit how
individuals produce and evaluate online self-presentations, and proposes that individuals
tend to rely on or value information with greater warranting value (Walther & Parks,
2002). Warranting value initially pertains to the match between an online self and reality;
individuals would perceive a warrant in a continuous fashion (DeAndrea, 2014). Walther
(2011) and his colleagues later broaden warranting values to any cue that cannot be easily
manipulated by human agents. Three types of sources online serve as warranting value,
including other-generated cues (e.g., Friends’ comments), self-generated cues (e.g., uploaded posts from user themselves), and system-generated cues (e.g., users’ ratings of a product), are outlined as below.
Self- and other-generated sources as warranting value. Warranting value is a psychological construct, reflecting perceptions about the extent to which information is
20
immune to its source’s manipulation (DeAndrea, 2014). Information providing warrants
can be derived from a variety of different online sources (Walther & Parks, 2002).
Walther and Parks (2002) suggest that other-generated cues could serve as warrants when
online users access a target’s social networks. These warranting values are contingent
upon the perceptions of a target’s social relationships and social environment in the real
world; it is also determined by the willingness of this target’ friends to rebuke deception.
Individuals can access other-generated cues on Twitter by viewing the comments or
replies from a target’s tweets.
Besides other-generated information online, self-generated cues and direct communication could contribute to the development of warrants (Walther & Parks,
2002). For instance, the profile pages on Twitter contain several verifiable elements, including individuals’ profile avatar and professional affiliations, as well as their uploaded posts, daily photos, and videos. Direct communication between online users can also influence credibility perceptions as online interactions may enhance stability and reduce uncertainties (Antheunis, Valkenburg, & Peter, 2010). Empirical evidence indicates that comments by friends on social network sites tend to affect perceptions of a user’s popularity (Hong, Tandoc, Kim, Kim, &Wise, 2012), physical attractiveness
(Walther et al., 2008), and social attractiveness (Antheunis & Schouten, 2011).
Previous research has been exploring the functions of self-generated and other- generated cues in online impression formation. For instance, Walther, Van der Heide,
Hamel, and Shulman (2009) examined the relative impact of self-generated versus friend- generated statements (i.e other-generated statements) about a target on the impressions receivers made by examining Facebook profiles and the wall postings on them. The
21
results support warranting theory and indicate that friends’ comments overrode self-
comments in perceiving physically attractiveness. Self-generated cues are considered to
attribute relatively low weight in warranting values compared to other sources of
information (Walther, 2011).
Although most scholars posit that social networking sites, such as Facebook and
Twitter, play a positive role in self-presentation and impression management, some argue
that online users are aware that online self-presentations are misleading among both
friends and acquaintances, which implies the conceptualization constraints and theoretical
boundaries of the warranting principle (DeAndrea, 2014; DeAndrea & Walther, 2011).
Although Walther et al.’s (2009) study found that friends’ comments overrode self-
comments in perceiving physical attractiveness, the study produced only ambiguous
support when measuring the perceptions about extraversion of the profile owner.
Similarly, Utz (2010) found mixed support when examining the effects of competing
claims between the self and the third party in impression formations on a social network site. The results demonstrate that self-claims tend to influence more on perceptions of the target’s popularity, whereas third-party claims had a greater effect on perceptions of the target’s communal orientation. DeAndrea (2014) explains that the co-presence of multiple warranting cues may mitigate or counteract any effects on perceptions of the warranting value. In order to increase the predictive and explanatory power of warranting theory, it is critical to explicit how to control different warranting values from different online sources to work together within certain social contexts to influence impression formation, especially for source credibility or general credibility judgment during risk and crisis situations.
22
System generated sources and information aggregates. Other than self- and
other-generated sources, system-generated cues can be significant determinants of social judgments on a target user that contribute to warranting value (Utz, 2010). Walther and
Jang (2012) indicate that system-generated information is robust for impression formation and evaluations in online settings, specifically because it is aggregate user- representations compiled and presented by computers or social networking sites, reflecting a collection of users’ ratings, votes, or behaviors. System-generated cues such as the ratings of a product, the number of friends or followers, as well as the geological locations and the time of the posts may enhance credibility perceptions of the product or the message content online. For instance, the number of friends indicated on a Facebook user’s profile page (a system-generated cue) would tend to influence perceptions of the users’ social attractiveness (Tong, Van Der Heide, Langwell, & Walther, 2008; Utz,
2010). The number of followers and the number of follows on Twitter (another system
generated cue) would work together to impact perceived credibility of Twitter users
(Westerman, Spence, & Van Der Heide, 2012).
Compared to information created online by human agents, information presented by computer systems or media affordances may be less likely to be manipulated or more difficult to be manipulated, which in turn may cause them to be perceived as having higher levels of warranting value (Antheunis & Schouten, 2011; Walther & Jang, 2012).
Furthermore, as the number of people contributing information expands, the warranting value of aggregate user-representations may increase in online environments. To be specific, given the number of retweets is a compilation of ratings from multiple sources
(users) indicated by the Twitter system, it is possible that online users would perceive
23
more stable warrants in credibility from retweets than from self-generated or other
generated information. Consistently, another online information evaluation theory,
signaling theory (Donath, 1999; 2007) also accounts for the conditions when individuals
would perceive online information to be privileged.
Signaling Theory
Signaling theory (Donath, 1999, 2007) posits that certain signals or information
are reliable for accessing information quality and source expertise. Signals refer to
“perceivable features and actions that indicate the presence of those hidden qualities,”
such as consumption patterns or the users’ statements (Donath, 2007, p. 233). Consistent
with warranting value, signaling theory applies cost-benefit analyses to specify what makes signals reliable and how people evaluate those signals (Donath, 2007). Two types of signals are outlined, these include, assessment signals and conventional signals, which are identified in terms of their nature and their variations in reliability. Assessment signals are inherently reliable and tethered to the quality they reflect, such as the price tag for a product; whereas conventional signals, such as self-descriptions in online profiles, vary in their reliability and are arbitrary indicators of some quality or trait (Wiley, 1983;
Hurd & Enquist, 2005). According to signaling theory, assessment signals are more reliable than conventional signals; the more that the costs of deception outweigh the
benefits of producing the signal, the more reliable the signal is perceived to be (Donath,
2007).
Signaling theory was initially developed in economics and biology, addressing a
variety topics such as the functioning of labor markets or animal behaviors, then the
theory was extended to the evaluation of online information (Donath, 2007). Previous
24 studies suggest that a variety of different signals could predict important online interaction outcomes. For instance, Shen, Chiou, and Kuo (2011) found that seller reputation, product condition, and argument quality predicted a number of bids, auction success, and willingness to pay in online auctions. Research on the judgments of strangers’ expertise online also suggests that online users place a larger emphasis on signals that are costly to fake when forming impressions of others (Shami, Ehrlich, Gay,
& Hancock, 2009).
Although convention signals are less reliable, there are more conventional signals such as “self-claims” predominantly available in online settings (Donath, 2007). The production and evaluation of conventional signals can be clarified through the analysis of costs and benefits (Donath, 2007). For instance, Donath (2007) outlines the costs associated with adding friends and evaluating profiles affect the reliability of users’ self- presentation. The study also examines strategies such as information fashion and risk- taking; it outlines how these costs and strategies affect the publicly-displayed social network in aiding the establishment of trust, identity, and cooperation. When evaluating online conventional signals, individuals have to assess the presence of costs and benefits for both senders and receivers (DeAndrea, 2014). Taking the number of retweets as an example, it takes time for multiple online users to review and decide to retweet an original tweet; people who retweet the post are also message receivers, making deception more difficult and costly. The retweet number is computed by the system and difficult to manipulate. Thus, the number of retweets is considered to be a reliable signal not only because the cost of deceptions in retweet numbers that outweigh the benefits for producing the number, but also for the reason that interactions with other users in a
25 target’s social network would increase source credibility perception of the target
(Walther, 2011).
The perspectives of signaling theory suggest that specific information cues may be particularly credible online (Flanagin & Metzger, 2013). In addition, Donath (2007) suggests that “the costs and benefits that motivate people to act in certain ways are in continuous flux” (p. 234). However, it is unclear when and which costs and benefits are salient for information production and evaluation within specific online settings.
Although both signaling theory and warranting theory elucidate how and when, for a source to exert control over information, it is unclear how different sources works together in guiding information evaluations. To address the above-mentioned questions, the MAIN model (Sundar, 2008) is applied as the dominant theoretical framework in the current study.
The MAIN Model
The MAIN Model (Sundar, 2008) takes a heuristic approach and posits that technological features transmit their own cues that may influence users’ perceptions of credibility by merely their presence on a media interface. These cues are embedded in broad technological affordances presented in most social media, including modality, agency, interactivity, and navigability (MAIN), to trigger cognitive heuristics about the nature of the underlying content (Sundar, 2008). Given the nature of online interactions with and through digital media, people would not assess credibility by solely considering the source of information. Instead, all available online interface cues, such as sources, message, and the medium, could trigger heuristics and provide mental shortcuts for
26 credibility assessments about the source (Sundar, 2008). Such interface cues as outlined in the MAIN Model work well with signaling theory and warranting theory.
Four broad categories of media cues. To systematically understand the effects of social media on online credibility assessments, Sundar (2008) identifies the relevant affordances offered by the technology and the relevant heuristics cued by the affordances.
Four broad categories of cues embedded in various technological affordances are outlined, including modality, agency, interactivity, and navigability cues. Although these cues embedded in the technological features are external to the media content, each of these cues could trigger several heuristics to affect online users’ evaluations on the content itself (Sundar, Jia, Waddell, & Huang, 2015).
Modality cues are associated with the structures of the media to trigger presence, novelty, intrusiveness, or distraction heuristics, which would influence users’ media experience relating to credibility assessment in an either positive or negative way
(Sundar, 2008). For instance, individuals may perceive more reality when information is conveyed through videos than through texts; information presented by three-dimensional visuals may be more likely to enhance learning performance than that presented by two- dimensional images (Kim, Park, & Sundar, 2012; Li, Zhang, Sundar, & Duh, 2013).
Individuals may yet perceive distractions when viewing a combined presence of multiple modalities or perceive less purchase intention when exposed to ads with fast animation speed compared to those with slower speed (Sundar & Kalyanaraman, 2004).
Consistently, Lachlan, Spence, Edwards, Reno, and Edwards (2014) examined the role of speed of updates on a Twitter feed with perceptions of trust. The results suggest an
27 indirect relationship where update speed drives elaboration, which in turn predicts perceptions of credibility.
Agency cues are associated with the perceptions of sourcing through online interactions, which would trigger various heuristics, such as authority, identity, bandwagon, machine, and social presence (Sundar, 2008). This category of cues reflect social media characteristics and afford a broad range of sourcing possibilities (Sundar,
2008). Kang (2010) analyzed how users assess credibility on blogs, and found that the source of a blog was the first component evaluated by the participants. The results suggest the factors, including the source’s knowledge, passion, transparency, reliability, and influence, would affect credibility judgments.
Agency cues are empowered not only by the users’ ability to serve as sources, but also embedded in salient interface agents and technological functions (Sundar et al.,
2015). Studies also show that online news stories may empower cues pertaining to competence or experience of the source. Online news stories authored by laypersons may trigger authority heuristics, which were evaluated as less credible than the same stories authored by doctors (Hu & Sundar, 2009). Another study on the number of followers on
Twitter found that too many or too few followers lowered perceptions of Twitter authors’ credibility (Westerman et al., 2012). Agency affordances can trigger a variety of potential heuristics, each of which can shape the perceptions of message content and the senders.
Interactivity cues are embedded in users’ interactivities and activities with new media, which could trigger a wide variety of cognitive heuristics such as responsiveness and telepresence. Sundar (2008) indicates interactivity as “the most distinctive affordances of digital media” (p. 85), and traditional media could present little
28
interactivity affordances. A variety of audience and situational factors may trigger
heuristic interactivity cues in a medium. Previous studies found incremental effects on learning when using virtual reality systems and immersive virtual environments (Wilson,
& D’Cruz, 2006).
Navigability cues could trigger various levels of heuristics when users are browsing, playing, or navigating digital environments. According to the MAIN model, the affordances of particular navigability tools on the interface, such as virtual simulations with an increased range of motion or search engines with relevant results, can
enhance user experience (Balakrishnan & Sundar, 2011; Kalyanaraman & Ivory, 2009).
Individuals may perceive the benevolence of the designers from navigability cues, and
are predisposed to positive opinions on the medium or message content (Sundar et al.,
2015).
The current set of studies focuses on agency cues and machine heuristics, to
explicate the effects of social media agents in influencing source credibility perceptions.
Given the growing popularity and penetration of media in risk and crisis situations, and
the continued ability of users to generate content without the filter of gatekeepers, issues
of how the public judge credibility in these environments merits further study. Types of
online agency heuristics embedded in retweet and user identities, including identity
heuristic and bandwagon heuristic, are particularly analyzed in the social media
environment of Twitter for the current set of studies. Also, in line with previous
arguments, identity heuristics and bandwagon heuristics are high in warranting value and
have high signal reliability.
Bandwagon Heuristics and the Numbers of Retweets
29
Bandwagon heuristics trigger credibility processing and employ the following logic: “if others think that this is a good story, then I should think so too” (Sundar, 2008, p. 83). Previous studies indicate that the perceived trustworthiness of a website or a source may be based on aggregated feedback (Flanagin & Metzger, 2008). Ample evidence indicates people utilize networks of peers and social computing applications for credibility evaluations, such as recommendation agents on Amazon or eBay. In spite of opinions about a product, individuals may be more reliant on collective comments and testimonials for purchase decision making (Flanagin & Metzger, 2008). Studies by
Sundar and Nass (2001) as well as Sundar, Knobloch-Westerwick, and Hastall (2007) both identified bandwagon heuristics as the most powerful cognitive shortcuts for evaluating online news, with the results that the news stories selected by other users received the highest ratings. Further, bandwagon cues with negative valance were revealed to influence cognitive processing more than the cues with positive valance
(Chevalier, & Mayzlin, 2006).
Researchers have identified bandwagon cues within machine heuristics on social media. For instance, Westerman et al. (2012) examined the impact of online cues available in social media on perceptions of a source’s credibility. Participants were guided to view Twitter profile pages with different numbers of followers and the ratios between followers and follows, and report their perceived source credibility. The results indicate having too many or too few network connections results in lower judgments of expertise and trustworthiness. Moreover, having a narrow gap between the number of followers and follows also led to increased judgments of competence (Westerman et al.,
2012).
30
The technological feature of showing the number of retweets embed in bandwagon cues is considered to be a unique attribute of the medium as well as its broad applicability in risk and crisis communication. The feature of retweets reflects non- reciprocal relationships in social media which is distinguished from offline interactions.
Meanwhile, as risk and crisis communication emphasizes the value of credibility evaluations, it is necessary to examine whether retweets at all matters for impression formation within these contexts. Consisted with the notions of warranting theory and signaling theory, bandwagon heuristic triggered by a number of retweets is compiled from aggregate user representations and system generations, which contains more constraints or costs for information deception. Thus, it is possible that online users would perceive greater source credibility when viewing the number of retweets. The following hypothesis is then proposed:
H: Higher levels of credibility will be assigned to the presence of a retweet
number in a risk information tweet than a tweet with the absence of retweet
number.
Identity Heuristic, and Interactions with Bandwagon
Another set of agency heuristics identified by Sundar (2008) is identity heuristics.
Identity heuristics are those that trigger credibility perceptions based on the allowance of users’ self-identity assertion through agency affordances on social media (Sundar, 2008).
Given information anonymity and aggregated authorship online, information associated with individual identities, such as the tweets posted by a peer on Twitter, might imply its content quality to a perceiver (Walther et al., 2010; Sundar, 2008) as compared to an unknown person. Additionally, message receivers who assess a peer’s identity cues, may
31
overlook their personal idiosyncrasies and focus instead on the perceived affiliation with
the peer (Lee, 2006). Thus, a tweet from a peer would enhance credibility appraisal to an
extent that would not be found in response to a tweet from a stranger or a person whom
the receiver cannot evoke identity heuristics.
Also, when tweets are posted by an expert on Twitter, the expert would serve as
authority source that may trigger authority heuristics and endorsement-based heuristics
pertaining to the quality and reliability of message content (Hilligoss & Rieh, 2008;
Sundar, Oeldorf-Hirsch, & Garga, 2008). Authority cues serve as “a major criterion for assigning credibility to a website” (Sundar, 2008, p. 84). Such cues assist the evaluation
of source credibility, especially in risk and health issues, given the inequality of
information flow between content producers and consumers. Previous research suggests
that individuals were more influenced by a medical expert or official authorities, such as
tweets generated by American Heart Association (Westerman et al., 2014). Thus, a tweet by a peer may have more value, than a tweet from a stranger, but an expert may be
viewed as more credible than any other identity.
Previous studies on authority and peer information sources during nation-wide
disasters have provided consistent support for their high credibility (e.g., Lachlan,
Spence, & Eith, 2014). Sundar (2008) has suggested that authority heuristics are
especially effective in impacting younger adults and youths who are more accustomed to
obedience. This research also suggests that younger adults would be influenced by peers
and group insiders (Sundar, 2008). The social networking and information aggregation on
social media provide potential for peer-to-peer credibility assessments, which may also
undermine traditional authorities (Flanagin & Metzger, 2008). It is, therefore, important
32 to re-evaluate the persuasive effects of authority and identity cues in social media.
Compared with legacy media, however, social media afford people multiple distributed authority cues instead of single authority based on scarcity and hierarchy, which enhanced authority heuristics (Madden & Fox, 2006). Given the high uncertainty and information inequity during risk and crisis situations, people would tend to use authority cues rather than peer cues as consistent and stable guidelines to influence decision making and credibility judgments (Sundar, Xu, & Oeldorf-Hirsch, 2009). Given that
Sundar (2008) found that authority cues are one of the most robust criteria in the credibility research literature, individuals may perceive a tweet from a risk and crisis practitioner more credible than from a peer (or any other identity cue with which the receiver has no affiliation).
However, mixed evidence has been indicated concerning the influence of bandwagon heuristics when placed alongside other identity heuristics. For instance,
Sundar et al. (2009) suggested that bandwagon and peer cues were more persuasive than authority cues for commercial decisions on Amazon; yet authority cues were more efficient when bandwagon or peer cues were inconsistent. There is also a dearth of data concerning the effects of bandwagon heuristic specifically in the context of Twitter. A study conducted by Lee and Sundar (2013) explored the source credibility of health messages in Twitter, and found that professional accounts with large numbers of followers were perceived to be more credible than a layperson with the same numbers of followers; whereas a lay person account with large numbers of followers retweeting a high-authority source was perceived to be less credible than a lay person account with the same numbers of followers retweeting a low-authority source. To provide a clear,
33
cohesive guideline for the source credibility perceptions in risk and health
communication, it is necessary to explore how bandwagon cues influence credibility
assessment in light of different user identities.
Moreover, previous research in warranting theory indicates the capability and
possibility that individuals manipulate computer generated information (Proile of the
2012 Presidential Candidates’ Twitter Followers, n.d). Thus, it is pragmatic to determine
the degree to which people view aggregate user representations and system-generated
cues as immune to manipulation (DeAndrea, 2014). Also, regarding signaling theory, less
specificity is offered regarding clarifying how the reliability of conventional signals varies (DeAndrea, 2014). Thus, it is necessary to explore to what extent the numbers of retweet would triggers cognitive processing as well. Therefore, the following research question is posited:
RQ 1: To what extent do identity and bandwagon heuristics differ in their ability
to induce source credibility in Twitter risk messages?
Source Credibility Dimensions and Measurement
Researchers have attempted to identify the underlying dimensions of source
credibility (Metzger & Flanagin, 2015). Scholars such as Gaziano and McGrath (1986)
have found competence and trustworthiness to be the most influential components in the
factor-analytic investigation of source credibility (O’Keefe, 1990). Priester and Petty
(1995), as well as McCroskey and Young (1981), indicate that trustworthiness and
competence tend to predict source acceptance and effectiveness. Consistently,
McCroskey and Teven (1999) further identify three primary interrelated dimensions of
source credibility, including perceived competence or expertise, trustworthiness, and
34
goodwill (McCrosky & Teven, 1999; O’Keefe, 1990). Several studies have used this
conceptualization (e.g., Edwards, Edwards, Spence, & Shelton, 2014; Lachlan et al.,
2014; Lim & Van der Heide, 2015; Limperos, Buckner, Kaufmann, & Frisby, 2015;
Omilion-Hodges & Rodriguez, 2014; Van der Heide & Lim, 2015; Westerman et al.,
2014).
Competence is regarded as the extent to which the source is perceived as qualified
or knowledgeable on a particular subject. Trustworthiness reflects an audience’s belief in
the integrity of a source; it refers to the extent to which the source is perceived to tell the
truth or to be reliable in only making assertions that the source truly believes. Whereas
goodwill refers to the extent to which the source is perceived as caring for the recipients
or having the best interests of the recipients at heart (Hovland, Janis, & Kelley, 1953;
Nimmo & Savage, 1976; Teven & McCroskey, 1997).
Some studies also found various secondary dimensions that may influence source credibility perceptions, such as caring, composure, dynamism, similarity, liking and sociability (Berlo, Lemert, Metz, 1969; Cronkhite & Liska, 1976; McCroskey & Teven,
1999; O’Keefe, 1990). Moreover, researchers, in turn, investigate the relative
effectiveness on each dimension of source credibility in attitude changes (Metzger et al.,
2003; Wilson & Sherrell, 1993). Although there is a debate about the precise factor
structure of source credibility (noted in Westerman et al., 2012), the construct with three
primary dimensions was commonly found in the credibility literature and has been widely
investigated in source credibility research, which is adopted in the two current studies.
It is necessary to incorporate the dynamic communication processes within both
individual and organizational levels to advance the understanding of communication
35
variance in risk and crisis processes. Organizational behavior literature indicates a micro- macro distinctions, consistent with the ideas of contextualization and multi-level research, which results in two parallel but largely non-intersecting disciplines of organizational science and sociology (Schneider, Goldstein, & Smith, 1995; Staw &
Sutton, 1992). By conducting contextualization, researchers make scientific references from a set of relevant, observable or measurable facts, and apply these references to form part of comprehensive knowledge (Cappelli, & Sherer, 1991). Given the complex, dynamic, interactive processes within organizational communication (for crisis and risk research), scholars conducting research at only one level of analysis may fall short in interpreting the true sources of variances (Hackman, 2003); whereas multi-levels of contextualization within organization research would increase the generalizability from one level to another. Therefore, Study Three attempts to begin the analysis of credibility perceptions that exist in both individual sources and organizational sources. To model the richness of organizational processes, is it critical to be able to examine at more than the individual level of source credibility and therefore Study Three is offered.
Researchers not only investigate credibility in the contexts of an individual’s public address, but also have expanded the original viewpoints to the areas of organizations and online media (Metzger & Flanagin, 2013; Sundar, 2008). Although some inconsistencies were found for the dimensions of source credibility in previous studies, such discrepancies indicate a necessity to measure source credibility perceptions within different types of sources and the contexts (Metzger & Flanagin, 2015). Moreover, it may be argued that the dimensions of source credibility used at the individual level are not appropriate for analysis at the organizational level. One measurement that may be
36 used for the use of credibility assessments at the organizational level is the RAND Public
Health Disaster Trust Scale (Eisenman et al., 2012).
Taking machine heuristics within risk and crisis contexts into consideration, the following research question is posited:
RQ 2: To what extent do the effects of bandwagon heuristics triggered by the
number of retweets also hold true for organizational level on source credibility
perceptions?
Chapter Summary
This chapter reviewed the literature on the roles of social media and credibility in risk and crisis communication. Guided by the theoretical framework of the MAIN model
(Sundar, 2008), warranty theory (Walther & Parks, 2002), and signaling theory (Donath,
1999; 2007), the literature review outlined how media affordances would influence source credibility assessments of health risk messages on a social media page. In particular, two types of machine heuristics, triggered by online media affordances, which would influence judgments of credibility on Twitter, including bandwagon and identity heuristics, were explored in the literature review. The chapter also proposed the need to examine the source credibility dimensions and measurements in risk and crisis studies.
The research hypothesis and the research questions within three independent studies were forwarded.
Copyright © Xialing Lin 2016
37
CHAPTER 3: METHODS
Full experiments strictly follow the rule of randomization when assigning treatments (Campbell & Stanley, 1963). This type of experimental designs refers to randomized experiments in which a random process, such as the toss of a coin or a table of random numbers, assigns units to receive the treatment or an alternative condition
(Shadish, Cook, & Campbell, 2002). In a full experimental design, participants are
randomly assigned to either the treatment or the control group, whereas they are not
assigned randomly in a quasi-experiment or a pre-experiment (Campbell & Stanley,
1963). In an experiment, the control and treatment groups differ not only in terms of the
experimental treatment they receive, but also in other, often unknown or unknowable,
ways. Thus, the researcher must try to statistically control for as many of these
differences as possible (Shadish et al., 2002). Because control is lacking in experiments
and pre-experiments, there may be several "rival hypotheses" competing with the
experimental manipulation as explanations for observed results (Shadish et al., 2002).
The full experiment is often considered as the only research methods that adequately
measure the cause and effect relationship (Shadish et al., 2002). To examine the
hypothesis and research questions in the current study, three post-test only full
experiments with self-report online surveys were conducted to examine how Twitter
affordances influence individuals’ source credibility perceptions in risk situations in the
following three studies. A fictional health risk and food safety topic was used as the risk
and crisis manipulation in the experiments.
38
Study One
The first experiment consisted of a two-condition, post-test only experiment with
a self-report online survey to examine whether retweets affect source credibility
assessment on Twitter. A two-condition design with retweets (present vs. absent) on a
mock stranger’s Twitter profile page was created. A stranger (person unknown to the
respondent) was used and all material not relevant to the manipulation was blurred out to
prevent experimental contamination. Participants were informed about the research
opportunity in class, and were directed to a website for the study. Participants first read
the informed consent; after choosing to continue, they were randomly assigned to either
one of the two experimental conditions. Respondents were asked to take their time to
view the page. After viewing the Twitter page, participants were directed to respond to a
questionnaire concerning their perceptions of source credibility and other demographic
questions. After this respondents were taken to a separate page to enter information for
extra credit. This process took no more than 15 minutes.
Participants
Participants were recruited as a convenient sample from undergraduate communication courses at a southern research university. Extra credit was offered in exchange for students’ participation. If students agreed, they would be sent an email with a link to the study. Confidentiality was assured by data collection without identifying information and a spate survey was used to collect information to award extra credit.
Materials
Participants were randomly assigned to view one of the two mock Twitter pages.
The Twitter pages were designed to display a user who posts a tweet regarding food
39 safety issue for each condition. In specific, the food safety issue was centered on the presence of contaminated watermelons in the market. This topic was chosen as it was a food safety rumor and health threats affecting broad audiences, which was relevant to the chosen sample. Also, the topic did not necessarily present immediate dangers to participants such as earthquakes or wildfires; participants would perceive the Twitter page as real without immediate updates.
Two mock Twitter pages provided retweet present-absent conditions with the same message content, sources, and updates. For the present condition, the tweet showed an amount of the retweet number (i.e., 400) underneath the original message. For the absent condition, the tweet contained no retweet number at all. All information other than the tweet about contaminated watermelons was blurred (see Appendix 1).
Measures
After viewing the mock Twitter page, the participants were guided to complete a posttest survey about source credibility. McCroskey and Teven’s (1999) source credibility scales with three separate constructs (i.e., competence, goodwill, and trustworthiness) were adapted. Each of the constructs contains six separate semantic differential type items, anchored with two antonyms (e.g., moral-immoral) and a seven- point response scale ranging from 1 to 7. Acceptable reliability was found in previous research in competence scale (e.g., α = .86; Westerman et al., 2014), goodwill scale (e.g.,
α = .76; Westerman et al., 2012), and trustworthiness scale (e.g., α = .84; Westerman et al., 2014).
Cognitive elaboration has, at times, been shown to mediate the relationship between credibility and various stimuli in previous studies (Westerman et al, 2014,
40
Spence et al, 2016); although was not hypothesized in this study, it was measured using a
version of Perse’s (1990) 5-item measure, modified to reflect the previously viewed
twitter page. Using a 5-point response scale (5 = strongly agree, 1 = strongly disagree),
people reported their level of agreement with each item (e.g., “When I looked at this
page, I thought about it over and over again”). Acceptable reliability of α = .76 was
found. For the demographic variables, including participants’ age, gender, race, income,
and Twitter use frequencies were measured as well.
Study Two
The second experiment intends to investigate the extent to which the numbers of retweets influences source credibility perceptions with the interactions of different individuals’ identity cues on Twitter. A 3 (the numbers of retweets of an original tweet:
40 vs. 400 vs. 4,000) × 3 (users’ identity cues: peer’s vs. expert’s vs. stranger’s profile
pages) post-test only experiment was conducted with an online survey. Participants were
informed about the research opportunity in class, and were directed to a website for the
study. Participants first read the informed consent; after choosing to continue, they were
randomly assigned to one of the nine experimental conditions. Respondents were asked
to take their time to view the page. After viewing the Twitter page, participants were
directed to respond to a questionnaire concerning their perceptions of source credibility
and other demographic questions. This process took no more than 15 minutes.
Participants
Participants were recruited as a convenient sample from undergraduate
communication courses at a southern research university. Extra credit was offered in
41
exchange for students’ participation. If students agreed, they would be sent an email with
a link to the study. Confidentiality was assured by data collection without identifying
information.
Materials
Participants were randomly assigned to view one of the nine mock Twitter pages.
The Twitter pages were designed to display the same food safety topic as that applied in
the first study. Inspired by the method designs in Westerman et al. (2012), three equally-
spaced numbers of retweets ranging across levels were selected to differ their effects (i.e.,
low: 40; moderate: 400; and high: 4,000). To presenting different types of identity cues,
an expert’s, a student’s, and a stranger’s Twitter profile pages were created. An
individual account of a food scientist working for the Food and Drug Administration
(FDA) serves as an expert to trigger authority heuristics, with clear authority identities
indicated on this expert’s profile page, including user’s name, the profile introduction,
organization affiliation, avatars, and background images. An individual account
representing a user who is studying at the same university as the participants was adapted, given that participants would perceive this user as an in-group peer. Moreover,
an individual account without any personal identities was created to represent a stranger.
Nine mock Twitter pages presented conditions with different retweet numbers and
different message sources. The first Twitter page represented a peer user posting the topic
of contaminated watermelons with low retweet numbers shown underneath this post. The
second Twitter page presented an expert from FDA posting the same topic with low
retweet numbers. The third Twitter page presented a stranger without any personal
identities posting the same topic with the same low retweet numbers. The fourth Twitter
42 page represented a peer user posting the same topic with moderate retweet numbers. The fifth Twitter page represented an expert from FDA posting the topic with moderate retweet numbers. The sixth Twitter page represented a stranger without any personal identities posting the topic with moderate retweet numbers. The seventh Twitter page represented a peer user posting the topic with high retweet numbers. The eighth Twitter page represented an FDA expert user posting the topic with high retweet numbers.
Moreover, the ninth Twitter page represented a stranger without any personal identities posting the topic with large number of retweets. Except for the account user’s identities and the number of retweets, all the other information such as the time of the updates, numbers of replies, followers, sidebar images, and the content of the posts were the same in these nine conditions.
Measures
After viewing the mock Twitter page, the participants were guided to complete a posttest survey about source credibility. McCroskey and Teven’s (1999) source credibility scales with three separate constructs (i.e., competence, goodwill, and trustworthiness) were adapted. Each of the constructs contains six separate semantic differential type items, anchored with two antonyms (e.g., moral-immoral) and a seven- point response scale ranging from 1 to 7. Acceptable reliability was found in previous research in competence scale (e.g., α = .86; Westerman et al., 2014), goodwill scale (e.g.,
α = .76; Westerman et al., 2012), and trustworthiness scale (e.g., α = .84; Westerman et al., 2014).
Cognitive elaboration was also measured using a version of Perse’s (1990) 5-item measure, modified to reflect the previously viewed Twitter page. Using a 5-point
43
response scale (5 = strongly agree, 1 = strongly disagree), people reported their level of
agreement with each item (e.g., “When I looked at this page, I thought about it over and
over again”). For the demographic variables, including participants’ age, gender, race, income, and Twitter use frequencies were measured as well.
Study Three
The third experiment intends to examine whether bandwagon heuristics invoked
by retweets would also influence Twitter users’ credibility assessment when the original
tweets were posted by an organizational account. A three-condition design with the
numbers of retweets (i.e., 40 vs. 400 vs. 4,000) was presented under an original tweet on
an organizational account’s profile page. Participants were informed about the research
opportunity in class, and were directed to a website for the study. Participants first read
the informed consent; after choosing to continue, they were randomly assigned to one of
the three experimental conditions. Respondents were asked to take their time to view the
page. After viewing the Twitter page, participants were directed to respond to a
questionnaire concerning their perceptions of source credibility and other demographic
questions. Respondents were then taken to a separate survey to enter extra credit
information. This process took no more than 15 minutes.
Participants
Participants were recruited as a convenient sample from undergraduate
communication courses at a southern research university. Extra credit was offered in
exchange for students’ participation. If students agreed, they would be sent an email with
44
a link to the study. Confidentiality was assured by data collection without identifying
information.
Materials
Participants were randomly assigned to view one of the three mock Twitter pages.
The Twitter pages were designed to display the same food safety topic as that applied in
the previous two experiments. Three equally spaced numbers of retweets ranging across
levels were selected to differ their effects (i.e., low: 40; moderate: 400; and high: 4,000).
These numbers were chosen to be consistent with Study Two and with previous research
(see Westerman et al., 2012). To present an organizational source identity, the FDA
official account was adapted for the Twitter user profile page with clear authority
identities, including the avatar, user introduction, and background images.
Three mock Twitter pages presented conditions with different retweet numbers on
an FDA’s Twitter profile page. The first condition represented the FDA official account
posting the topic of contaminated watermelons with low retweet numbers shown
underneath this post; the second condition represented the FDA official account posting
the same topic with moderate retweet numbers; and the third condition presented the
FDA official account posting the topic with high retweet numbers. Except the numbers of
retweets were different among the three conditions, all the other information such as the
account user’s information, time of the updates, numbers of replies, and the content of the
posts were the same.
Measures
One limitation of studies by Westerman et al. (2014) and Spence et al. (2016) are the use of an interpersonal scale to measure credibility at the organizational level. To
45
replicate and extend those findings and guarantee the measurement validity and reliability
in the study, Study Three adopts the RAND Public Health Disaster Trust Scale
(Eisenman et al., 2012) to in the experiment. Four items scored on a 4-point Likert scale
measure the construct of trust, ranging from very confident to not at all confident
(Eisenman et al., 2012). These items include: 1) “How confident are you that the FDA
can respond fairly to your health needs, regardless of your race, ethnicity, income, or
other personal characteristics”; 2) “how confident are you that the FDA provides honest
information to the public”; 3) “how confident are you that the FDA can response effectively to protect the health of the public”; 4) “if there were a health problem associated with watermelons the market, and the FDA needed to collect information from you, how confident are you that this information would not be used against you”.
Acceptable reliability was found, with α = .82. The RAND Public Health Disaster Trust
Scale has been found to be highly correlated with the source credibility scale (see
Lachlan et al., 2014a).
Cognitive elaboration was also measured using a version of Perse’s (1990) 5-item
measure, modified to reflect the previously viewed Twitter page. Using a 5-point
response scale (5 = strongly agree, 1 = strongly disagree), people reported their level of
agreement with each item (e.g., “When I looked at this page, I thought about it over and
over again”).
As with Study One and Study Two, demographic variables, including
participants’ age, gender, race, income, and Twitter use frequencies were measured as
well.
46
Chapter Summary
This chapter outlined the research methods for the three studies. Study One aims to explore whether bandwagon effects triggered by the number of retweets would impact individuals’ source credibility evaluations. Study Two intends to examine the effects between different levels of retweet numbers based on three types of user identity cues.
Finally, Study Three investigates whether bandwagon effects triggered by the number of retweets would still hold true on credibility evaluations when user identity are in organizational level. Conceptual operationalization, participant recruitments, experiment designs, the implementation of the experiments, and measurement instruments in each study were outlined in the chapter.
Copyright © Xialing Lin 2016
47
CHAPTER 4: RESULTS
Study One
Sample Profile
A total of 111 valid responses from participants recruited from undergraduate classes at a large southern research university were included in Study One, the breakdown for biological sex included 45.9% males (n = 51) and 54.1% females (n = 60).
Most of the participants identified themselves as Caucasian (84.7%), followed by
African-American (7.2%), Asian (5.4%), and Latino (.9%); 1.8% of the participants identified their races as other. The majority of the respondents came from high and middle socio-economic levels, with 37.6% reporting annual household income over
$100,000, 38.5% between $50,001 and $ 100,000, and 8.3% below $20,000. Of these
participants, 66 reported their age (M = 18.61, SD = 1.23), ranging from 18 to 26 years
old (See Table 4.1).
The participants also reported their daily Twitter usage behaviors, with a majority
of the participants having Twitter accounts (75.7%). On average, the participants reported checking Twitter 12.7 times a day (SD = 15.12), spending 29.35 minutes daily on Twitter
(SD = 22.04, and following 351.83 users on their Twitter accounts (SD = 244.86, see
Table 4.1).
Confirmatory Factor Analysis on Source Credibility Scales
Although previous studies have indicated high reliabilities for McCroskey and
Teven’s (1999) scales, it could not assume that the high validations of these scales are across all subsequent uses (Levine, Hullett, Turner, & Lapinski, 2006). Levine et al.
(2006) advance an argument in favor of conducting and reporting confirmatory factor
48
analyses (CFA) on existing and previously validated scales, especially when there are
subtle changes in item wordings from study to study. To determine whether the data were
compatible with expected factors in McCroskey and Teven’s (1999) three-factor source
credibility scales, a CFA was conducted on the collected data using a maximum likelihood (MLM) solution in AMOS. After removing one item from goodwill scale (i.e.,
“self-centered; not self-centered”), the data was consistent with the three-factor solution and yielded good model fit indices: (116) = 229.43, CMIN/df = 1.98, TLI = .90, CFI 2 = .92, RMSEA = .09 (Levine, 2005).𝜒𝜒 High reliabilities were displayed in these three
scales after CFA, with α = .92 in competence, α = .89 in trust, and α = .87 in goodwill.
Bandwagon Heuristics and Source Credibility
The hypothesis predicted that people will perceive higher levels of credibility
with the presence of a retweet number in a risk information tweet than a tweet with the
absence of retweet number. To investigate the hypothesis, a series of independent sample
t-tests were conducted. Retweet number was treated as an independent variable, and three
dimensions of source credibility perceptions, including competence, goodwill, and
trustworthiness, were measured as dependent variables. As these dependent variables
were conceptually different, moreover, given the significant Box’s M test (p =. 031), the
dependent variables were analyzed in univariate level.
The results indicated that there were no significant differences between a tweet
with retweet numbers (M = 3.26, SD = 1.04) and a without retweet numbers (M = 3.26,
SD = 1.06) in perceiving competence, t (109) = .003, p = .997. There were no significant
differences between the condition with retweet numbers (M = 4.36, SD = 1.07) and
without retweet numbers (M = 4.48, SD = .93) in perceiving trustworthiness, t (109) =
49
-.339, p = .735. Finally, there were no significant differences between the condition with retweet numbers (M = 4.33, SD = 1.08) and without retweet numbers (M = 4. 48, SD =
1.01), t (109) = -.737, p = .463 in perceiving goodwill. Thus, the t-test results did not support the hypothesis.
Path Analyses
The aforementioned literature suggested a mediating effect of cognitive elaboration on the relationship between social media affordances and perceived source credibility (e.g., Westerman et al., 2014; Spence, Lachlan, Edwards, & Edwards, 2016).
Therefore, it is possible that cognitive elaboration would mediate the bandwagon effects triggered by the presence of retweet numbers on the three factors of the McCrosky and
Tevin (1999) credibility perceptions, namely competence, trustworthiness, and goodwill.
Path analysis was used to simultaneously test the interrelationships among the variables
(Bollen, 1989; Kline, 1998). Three simple thee-variable path analyses, using an MLM solution in AMOS, were applied to test the hypothesis involving mediation.
For competence, the results of the path analysis suggested a strong fit for the experimental data, (1) = .020, p = .888. Diagnostic statistics on the model fit further 2 confirmed the model𝜒𝜒 fit, CMIN/df = .020, CFI = 1.00, RMSEA = .000. The standardized
regression coefficient for the path between retweet numbers and cognitive elaboration
was .04 (p =.682), whereas the link between cognitive elaboration and perceived
competence was .32 (p <.001).
Similar strong results were detected for the indicators of trustworthiness and
goodwill. For trustworthiness, the results of the path analysis suggested a strong fit for
the experimental data, (1) = .030, p = .863. Diagnostic statistics on the model fit 2 𝜒𝜒 50
further confirmed the model fit, CFI = 1.00, RMSEA = .000. The standardized regression coefficients for the path between retweet numbers and cognitive elaboration was .04 (p
= .682; as it is in the remaining analyses), whereas the link between cognitive elaboration
and perceived trust was .45 (p < .001).
For goodwill, the results of the path analysis suggested a good fit for the
experimental data, (1) = .399, p = .528. Diagnostic statistics on the model fit further 2 confirmed the model𝜒𝜒 fit, CMIN/df = .399, CFI = 1.00, RMSEA = .000. The standardized
regression coefficient for the path between retweet numbers and cognitive elaboration
was .039 (p =.682; as it is in the remaining analyses), whereas the link between cognitive
elaboration and perceived goodwill was .38 (p < .001).
Demographics and Twitter Usage
To investigate whether participants’ demographic variables and their Twitter daily
usage have effects on credibility, three series of stepwise regression analyses were
conducted. Three factors of source credibility, namely competence, trustworthiness, and
goodwill, served as the three dependent variables separately. Each series of regression
analyses performed three predictor blocks. On the first predictor block, demographic
variables, including gender, age, and income were entered into the predictor block
(assumptions of linearity, normally distributed error, and uncorrelated errors were
checked and met). The second predictor block contained participants’ Twitter daily usage
measurements, including possession of Twitter account, frequency and length of their
daily use, and the number of users they followed (assumptions of linearity, normally
distributed error, and uncorrelated errors were checked and met). The third predictor
block contained the experimental condition of retweet number as a predictor.
51
For competence perception, results of the first predictor block indicated a
significant model, F (3, 101) = 5.68, p = .001, adjusted R2 = .119. Of the demographic
variables in the model, only gender (β = .727, p < .001) were predictive of source
competence. When Twitter daily usage measurements were added to the predictor block,
a significant model was produced, F (7, 97) = 3.46, p = .002, adjusted R2 = .142. Of the
demographic and Twitter daily usage variables in the model, only gender (β = .718, p
= .001) were predictive of source competence. When experimental condition was added
to the predictor blocks, a significant model was also produced, F (8, 96) = 3.00, p = .005, adjusted R2 = .134. Consistently, only gender (β = .715, p = .001) were predictive of
source competence of all the variables in the model. Thus, the regression analyses
suggest that only gender predicted participants’ competence perceptions.
For trustworthiness perception, results did not indicate a significant model for the
first predictor block, F (3, 101) = 1.64, p = .184, adjusted R2 = .018; nor for the second
predictor block, F (7, 97) = 1.28, p = .270, adjusted R2 = .018; nor for the third predictor
block, F (8, 96) = 1.11, p = .366, adjusted R2 = .008. Thus, participants’ demographics or
their daily Twitter usage did not predict their trustworthiness perceptions.
For perceptions of goodwill, consistently, results again did not indicate a
significant model for the first predictor block, F (3, 101) = 1.10, p = .352, adjusted R2
= .003; nor for the second predictor block, F (7, 97) = .845, p = .553, adjusted R2 = -.011;
nor for the third predictor block, F (8, 96) = .827, p = .581, adjusted R2 = -.013. Thus, participants’ demographics or their daily Twitter usage did not predict their goodwill perceptions.
52
Study Two
Sample Profile
A total of 434 valid responses were collected from undergraduate classes at a large southern research university for Study Two, with 45.9% males (n = 199) and 53.3%
females (n = 232). Of these participants, 331 reported their ages (M = 20.34, SD = 5.96), ranging from 18 to 55 years old. Most of the participants identified themselves as
Caucasian (79.0%), followed by Asian (7.4%), African-American (7.1%), Latino (3.0%),
and other races (3.0%). The majority of the respondents came from high and middle socio-economic levels, with 42.9% reporting annual household income over $100,000,
32.9% between $50,001 and $ 100,000, and 10.1% below $20,000 (See Table 4.2 for demographics).
The participants also reported their daily Twitter usage behaviors, with a majority of the participants having Twitter accounts (82.9%). On average, these participants reported checking Twitter 16.49 times a day (SD = 19.60), spending 31.64 minutes daily on Twitter (SD = 26.66), and following 339.19 users on their Twitter accounts (SD =
225.57; see Table 4.2).
Confirmatory Factor Analysis on Source Credibility Scales
To determine whether the data were compatible with the three expected factors in
McCroskey and Teven’s (1999) source credibility scales, a confirmatory factor analysis
(CFA) was conducted on the collected data using a maximum likelihood (MLM) solution in AMOS. Although previous studies have indicated high reliabilities for McCroskey and
Teven’s (1999) scales, it could not assume that the high validations of these scales are across all subsequent uses (Levine, Hullett, Turner, & Lapinski, 2006). Levine et al.
53
(2006) advance an argument in favor of conducting and reporting CFA on existing and previously validated scales, especially when there are subtle changes in item wordings from study to study. After removing one item from the goodwill scale (i.e., “not understanding-understanding”), the data was consistent with the three-factor solution and yielded good model fit indices: (111) = 223.81, CMIN/df = 2.11, TLI = .98, CFI 2 = .98, RMSEA = .05 (Levine, 2005)𝜒𝜒 . High reliabilities were displayed in these three scales after CFA, with α = .94 in competence, α = .93 in trust, and α = .87 in goodwill.
Source Credibility Perceptions
Study Two investigated the impact of identity and bandwagon heuristics on individuals’ source credibility perceptions of Twitter risk messages, namely competence, goodwill, and trustworthiness. As these three factors of source credibility were conceptually different, the current study examined univariate effects to investigate each underlying factors. To test the research question, three 3 (Twitter user identity: Expert vs.
Peer vs. Stranger) × 3 (Levels of retweet numbers: 40 vs. 400 vs. 4,000) analyses of variance (ANOVAs) were conducted to compare the effects of user identity and retweet number and their interactions on participants’ perceptions of source credibility regarding a Twitter risk message. Twitter user identity and retweets number were treated as independent variables; competence, trustworthiness, and goodwill were measured as three continuous dependent variables. The statistics with Tukey posthoc tests for each dependent variable are reported as below.
Competence. The ANOVA produced no significant interaction effect between user identity and retweet number on competence, F (4, 425) = .21, p = .932, suggesting no different impacts in sense of Twitter user identity depending on the levels of retweet
54
numbers in perceiving source competence. Table 4.4.1 reports the statistic descriptions
for independent variables on competence.
A significant main effect of user identity was found on competence perceptions, F
(2, 425) = 32.97, p <.001, = .13, Power = 1.00. Participants viewing an FDA expert’s 2 𝑝𝑝 Twitter account (M = 5.08,𝜂𝜂 SD = 1.24) were significantly more likely to perceived higher
levels of source competence than those viewing a peer’s (M = 4.20, SD = 1.07) and a
stranger’s (M = 4.01, SD = 1.12). There were no significant differences between a peer’s
and a stranger’s Twitter page in perceiving competence (see Table 4.4.2).
The results also indicated a significant main effect on retweet number, F (2, 425)
= 4.37, p =.013, = .02, Power = .754. Participants perceived the highest levels of 2 𝑝𝑝 source competence𝜂𝜂 when viewing the risk post with 40 retweets (M = 4.55, SD = 1.26),
followed by those viewing the post with 400 retweets (M = 4.46, SD = 1.22), and those
with 4,000 retweets (M = 4.14, SD = 1.14). In addition, Tukey post-hoc test indicated
significant differences between the risk post with 4,000 retweets and that with 40 retweets in source competence (see Table 4.4.3).
Trustworthiness. The ANOVA produced no significant interaction effect between user identity and retweet number on trustworthiness, F (4, 425) = .15, p = .964,
suggesting no different impacts in sense of Twitter user identity depending on the levels of retweet numbers in perceiving source trustworthiness. Table 4.5.1 reports the statistic descriptions for independent variables on trustworthiness.
A significant main effect of user identity was found on trustworthiness, F (2, 425)
= 4.67, p = .010, = .02, Power = .784. Consistently, participants viewing an FDA 2 𝑝𝑝 expert’s Twitter account𝜂𝜂 (M = 4.80, SD = 1.20) were significantly more likely to
55
perceived higher levels of source trustworthiness than those viewing a peer’s (M = 4.47,
SD = 1.01) and a stranger’s (M = 4.44, SD = .92). There were no significant differences
between a peer’s and a stranger’s Twitter page in perceiving trustworthiness (see Table
4.5.2).
The results also indicated a significant main effect of retweet number on
trustworthiness, F (2, 425) = 3.58, p = .029, = .02, Power = .663. Participants 2 𝑝𝑝 perceived the highest levels of trustworthiness𝜂𝜂 when viewing the risk post with 400
retweets (M = 4.68, SD = 1.02), followed by those viewing the post with 40 retweets (M
= 4.62, SD = 1.01), and those with 4,000 retweets (M = 4.36, SD = 1.01). In addition, there were significant differences between the risk post with 4,000 retweets and that with
400 retweets in trustworthiness (see Table 4.5.3).
Goodwill. The ANOVA yielded no significant interaction effect, F (4, 425) = .87, p = .964, or main effect for retweet number on goodwill, F (2, 425) = .63, p = 534. Thus,
retweet numbers do not differ in their levels in perceiving source goodwill despite
various user identities. Table 4.6.1 reports the statistic descriptions for independent
variables on competence.
However, a significant main effect of user identity was found on goodwill, F (2,
425) = 9.38, p < .001, = .04, Power = .978. Participants viewing an FDA expert’s 2 𝑝𝑝 Twitter account (M = 4.60,𝜂𝜂 SD = 1.13) were significantly more likely to perceived higher
levels of source goodwill than those viewing a stranger’s (M = 4.12, SD = 1.15) and a
peer’s (M = 4.08, SD = 1.00). There were no significant differences between a peer’s and
a stranger’s Twitter page in perceiving goodwill (see Table 4.6.2).
56
Study Three
Sample Profile
A total of 169 valid responses were obtained from participants recruited from undergraduate classes at a large southern research university for Study Three, with 33.1% males (n = 56) and 66.3% females (n = 112). Most of the participants identified themselves as Caucasian (72.8%), followed by Asian (13.0%), African-American (8.3%), and Latino (1.8%); 4.1% of the participants identified their races as others. Of these participants, 137 reported their ages (M = 19.67, SD = 1.46), ranging from 18 to 25 years old. The majority of the respondents came from high and middle socio-economic levels, with 33.7% reporting annual household income over $100,000, 21.3% between $70,001 and $100,000, 19.5% between $50,001 and $70,000, 17.7% between $20,001 and
$50,000, and 7.7% below $20,000 (see Table 4.3).
The participants also reported their daily Twitter usage behaviors, with a majority of the participants having Twitter accounts (74.0%). On average, the participants reported checking Twitter 15.70 times a day (SD = 18.37), spending 32.30 minutes daily on
Twitter (SD = 29.81), and following 318.57 users on their Twitter accounts (SD = 212.88; see Table 4.3).
Source Credibility Perceptions on Organizational Level
Study Three proposed a research question inquiring whether the effects of bandwagon heuristics triggered by the number of retweets would also hold for source credibility perceptions at the organizational level, specifically using the RAND Public
Health Disaster Trust Scale. To test the research question, the current study conducted a one-way ANOVA with Tukey posthoc to evaluate the variability in credibility
57 perceptions across three conditions. The analysis found that there were significant differences among the conditions for source credibility perceptions, F (2, 166) = 3.13, p
= .046, = .04, Observed Power = .60 (see Table 4.7). 2 p Theη results on this reverse coded scale revealed that the participants who viewed the FDA Twitter page containing 4,000 retweets under the risk information post (M =
2.13, SD = .60) were significantly more likely to perceive lower source credibility than those viewed the FDA Twitter page containing 40 retweets under the same post (M =
1.86, SD = .47). However, there were no significant differences to compare the condition of 400 retweets (M = 1.95, SD = .58) with the condition of 4,000 retweets, nor with the condition of 40 retweets. Thus, the results suggest that higher levels of bandwagon heuristics embedded in retweet numbers would trigger lower levels of source credibility perceptions towards an organization (see Table 4.7).
Chapter Summary
This chapter analyzed the data and reports the results in each study. For Study
One, the hypothesis was not directly supported. For Study Two, the data did not indicate significant interaction effects between user identity and retweet number on each dimensions of source credibility perceptions. Individual expert Twitter accounts were viewed to be the most credible among all the three identities. However, participants perceived the lowest levels of source credibility when viewing the conditions of high level retweet number. For study three, the data reveals consistency regarding a reverse bandwagon effect on credibility perceptions when user identities are presented on organizational level.
Copyright © Xialing Lin 2016
58
Table 4.1 Study One: Demographic Characteristics of Participants
N (%) M SD Sex Male 51 (45.9) Female 60 (54.1) Age (18-26) 62 18.61 1.23 Race Caucasian 94 (84.7) African-American 8 (7.2) Asian 6 (5.4) Latino 1 (.9) Others 2 (1.8) Income Under $20,000 9 (8.1) $20,000 - $30,000 2 (1.8) $30,001 - $50,000 15 (13.5) $50,001 - $70,000 22 (19.8) $70,001 - $100,000 20 (18.0) Over $ 100,000 41 (36.9) Twitter Account Ownership Yes 84 (75.7) No 21 (18.9) Twitter Daily Usage Frequency Checking Twitter (times/day) 12.7 15.12 Length on Twitter 29.35 22.04 (min/day) Users Following on Twitter 351.83 244.86 N 111
59
Table 4.2
Study Two: Demographic Characteristics of Participants
N (%) M SD Sex Male 199 (45.9) Female 232 (53.5) Age (18-55) 330 20.34 5.96 Race Caucasian 343 (79.0) African-American 31 (7.1) Asian 32 (7.4) Latino 13 (3.0) Others 13 (3.0) Income Under $20,000 44 (10.1) $20,000 - $30,000 26 (6.0) $30,001 - $50,000 31 (7.1) $50,001 - $70,000 63 (14.5) $70,001 - $100,000 80 (18.4) Over $ 100,000 186 (42.9) Twitter Account Ownership Yes 360 (82.9) No 73 (16.8) Twitter Daily Usage Frequency Checking Twitter (times/day) 351 16.49 19.60 Length on Twitter (min/day) 345 31.64 26.66 Users Following on Twitter 320 339.19 225.57 N 434
60
Table 4.3
Study Three: Demographic Characteristics of Participants
N (%) M SD Sex Male 56 (33.1) Female 112 (66.3) Age (18-25) 137 19.67 1.46 Race Caucasian 123 (72.8) African-American 14 (8.3) Asian 22 (13.0) Latino 3 (1.8) Others 7 (4.1) Income Under $20,000 13 (7.7) $20,000 - $30,000 11 (6.5) $30,001 - $50,000 19 (11.2) $50,001 - $70,000 33 (19.5) $70,001 - $100,000 36 (21.3) Over $ 100,000 57 (33.7) Twitter Account Ownership Yes 125 (74.0) No 44 (26.0) Twitter Daily Usage Frequency Checking Twitter (times/day) 15.7 18.37 Length on Twitter (min/day) 32.30 29.81 Users Following on Twitter 318.57 212.88 N 169
61
Table 4.4.1
Study Two: Descriptive Statistics for Competence as a Function of User Identities and Retweet Numbers
FDA expert Peer Stranger (No Identity) Retweet Numbers n M SD n M SD n M SD 4,000 37 4.73 1.22 45 4.04 1.07 50 3.79 .99 400 44 5.20 1.18 49 4.22 1.04 51 4.04 1.15 40 49 5.22 1.29 56 4.32 1.10 53 4.17 1.18 Note: N =434.
Table 4.4.2
Study Two: Tukey Post-hoc Test Multiple Comparisons for User Identity on Competence
User Identity M (SD) FDA Expert Peer Stranger FDA Expert 5.08 (1.24) *** *** Peer 4.20 (1.07) *** Stranger 4.01 (1.12) *** Note: F (2, 425) = 32.97, p < 001, = .13, Power = 1.00. ***p < .001. 2 𝑝𝑝 𝜂𝜂 Table 4.4.3
Study Two: Tukey Post-hoc Test Multiple Comparisons for Retweet Numbers on Competence
Retweet M (SD) 4,000 400 40 Number 4,000 4.14 (1.14) *** 400 4.46 (1.22) 40 4.55 (1.26) *** Note: F (2, 425) = 4.37, p =.013, = .02, Power = .754. ***p < .001. 2 𝑝𝑝 𝜂𝜂
62
Table 4.5.1
Study Two: Descriptive Statistics for Trustworthiness as a Function of User Identities and Retweet Numbers
FDA expert Peer Stranger (No Identity) Retweet Numbers n M SD n M SD n M SD 4,000 37 4.53 1.11 45 4.33 1.04 50 4.26 .89 400 44 4.98 1.14 49 4.55 1.02 51 4.54 .85 40 49 4.84 1.29 56 4.52 .99 53 4.52 1.00 Note: N =434.
Table 4.5.2
Study Two: Tukey Post-hoc Multiple Comparisons for User Identity on Trustworthiness
User Identity M (SD) FDA Expert Peer Stranger FDA Expert 4.80 (1.20) * ** Peer 4.47 (1.01) * Stranger 4.44 ( .92) ** Note: F (2, 425) = 4.67, p = .010, = .02, Power = .784. **p =. 010, * p < .05. 2 𝑝𝑝 𝜂𝜂 Table 4.5.3
Study Two: Tukey Post-hoc Multiple Comparisons for Retweet Numbers on Trustworthiness
Retweet Number M (SD) 4,000 400 40 4,000 4.36 (1.01) * 400 4.68 (1.02) * 40 4.62 (1.10) Note: F (2, 425) = 3.58, p = .029, = .02, Power = .663. *p < .05. 2 𝑝𝑝 𝜂𝜂
63
Table 4.6.1
Study Two: Descriptive Statistics for Goodwill as a Function of User Identities and Retweet Numbers FDA expert Peer Stranger (No Identity) Retweet Numbers n M SD n M SD n M SD 4,000 37 4.56 1.04 45 4.10 1.05 50 3.90 1.12 400 44 4.67 1.13 49 4.07 1.03 51 4.09 1.03 40 49 4.56 1.21 56 4.07 .96 53 4.36 1.25 Note: N =434.
Table 4.6.2
Study Two: Tukey Post-hoc Test Multiple Comparisons for User Identity on Goodwill
User Identity M (SD) FDA Expert Peer Stranger FDA Expert 4.60 (1.13) *** ** Peer 4.08 (1.00) *** Stranger 4.12 (1.15) ** Note: F (2, 425) = 9.38, p < .001, = .04, Power = .978. ***p < .001. ** p = .001 2 𝜂𝜂𝑝𝑝
Table 4.7
Study Three: ANOVA and Tukey Post-hoc Test Results among Three Conditions
F df Power R2 Credibility 3.13 2, 166 .0462 .60 .04 η𝑝𝑝 Conditions M (SD)+ N Con 1 Con 2 Con 3 Con 1: FDA with 4k retweets 2.13 (.60) 50 * Con 2: FDA with 400 retweets 1.95 (.58) 60 Con 3: FDA with 40 retweets 1.86 (.47) 59 * Note: * p <.05. Con = Condition. + RAND is a reverse coded scale.
64
CHAPTER 5: DISCUSSION
General Discussion
Recognizing that risk and crisis theories have largely neglected the capacity of social media affordances in inducing effective responses and crisis resilience to health risk appeals, this research sought to use and apply theory to examine the functional roles of diverse media affordances in online health risk cognitive processes. Considering that source credibility serves as a critical persuasive attribute for risk information acceptance and interpretation, these studies focus on source credibility evaluations within a health risk forum. Media research, as well as risk and crisis scholars, argue that diverse media affordances play critical roles in helping people perceive source credibility, as they construct and reinforce both systematic and heuristic information processing, especially in risk and crisis circumstances.
Limited research has directly examined the effects of bandwagon and identity heuristics on social media in risk and health contexts. These studies draw from the MAIN model (Sundar, 2008), warranting theory (Walther & Parks, 2002), and signaling theory
(Donath, 1999; 2007), and proposed that retweet number and user identity would trigger diverse cognitive heuristics for source credibility judgments in a health risk topic.
Specifically, the current research contains three studies. Study One found no direct relationship between retweet number and each of the dimensions of source credibility.
Study Two found that authority identity may play a significant role in influencing credibility perceptions; a resistance effect or reverse bandwagon effect for different levels of retweet numbers was revealed from the data. Study Three, compared with Study Two, took a step further and found a consistent reverse bandwagon effect for retweet numbers
65
when sources are from both the individual level and organizational level. The findings
suggest several important theoretical and practical implications, which are discussed as
below.
Study One
Study One used an experimental design to examine whether retweet numbers
under a health risk post would have an impact on viewers’ source credibility perceptions.
In particular, participants viewed a tweet about the presence of contaminated watermelons in the markets. The retweet numbers were manipulated, and a stranger’s
Twitter page without any personal identities was used; information other than the content
of the tweet was blurred. Participants responded to a measure of cognitive elaboration
(Perse, 1990) and source credibility (McCroskey & Teven, 1999), as well as reporting the
information on their demographics and daily Twitter usage. The results revealed that the number of retweets, when presented alone, did not impact source credibility; but rather, cognitive elaboration had a role in mediating the relationship of retweet number and the
perceptions of source credibility. The findings harken back to previous research of
heuristic cues on online information processing (e.g., Spence et al., 2016; Westerman et
al., 2014), and extended previous understandings surrounding the functions of diverse
social media affordances in risk and health contexts.
Primary Findings
Retweet number did not directly influence participants’ source credibility. The results of t-tests indicated no differences between the condition with retweets and that
without retweets. One possible explanation would be related to participants’ defensive
66 processing of health risk information (Witte, Meyer, & Martell, 2001). Participants might tend to process the tweet of the prevalence of contaminated watermelons through the central processing route, which relied less on bandwagon heuristics triggered by retweet numbers.
The manipulated food safety topic in the experimental design is relevant to participants’ daily life; unlike H1N1 influenza or terrorist attacks which require higher levels of professional and technical ability to take self-protective actions, participants might perceive more efficacy to react to this health risk issue. In turn, the information might be processed in an adaptive manner (Witte, 1992). Also, in both conditions, participants were only exposed to a single post from a stranger, which avoids the situation of information overload. Participants might tend to use their pre-existed knowledge and experience to process the information without applying bandwagon heuristics.
Consistently, previous research on elaboration likelihood suggests that individual would tend to process health risk messages centrally only when the messages are important and relevant to them, and when individuals have the ability to process the messages (Petty & Cacioppo, 1986; Witte et al., 2001). The t-tests results showed that, regardless of the presence of retweets, participants tend to perceive the source as low in competence, but moderate in trustworthiness and goodwill pertaining to the food risk message. Thus, it is possible that participants did not adapt retweet number to their source credibility processing.
However, the path analysis models suggest that cognitive elaboration might moderate the relationship between retweet numbers and source credibility. The data were
67 not consistent with the linear predictions for the impact of retweet number on source credibility perceptions. Notably, the results predicted positive relationships between cognitive elaboration and each of the individual factors of source credibility. Significant correlations were found between cognitive elaboration and competence, goodwill, and trustworthiness. The findings were consistent with the mediation effects of machine heuristics during information processing in Spence et al. (2016) and Westerman et al.
(2014). When individuals involved in more cognitive elaboration, they tend to perceive stronger source credibility on the target user.
The overall robust model fit of path analyses also suggest that, although retweet number presented in the manipulation could not directly trigger different source credibility perceptions compared to condition without retweet number, it cannot be concluded that retweet number do not assist credibility judgments in all information processing situations. It is possible that there was a threshold to trigger source judgments when a certain amount of retweet numbers is displayed. This is, of course, speculative and more research focusing on presence and absence of various bandwagon cues needs to be conducted.
Limitations and Future Directions
Limitations of Study One arise from some features of the experimental design.
First, the study used only one tweet in the manipulation. This was done to attempt to avoid distractions from the experimental manipulation. Although other studies have used this technique, it does decrease ecological validity.
Similarly, the experiment used a static Twitter page for each condition which also may decrease ecological validity. Similar to previous social media affordances studies, a
68
captured screenshot of the mock Twitter pages were presented to ensure the internal validity of the experiment. Such designs ran short in providing “live running” nature of an actual Twitter feed, which could not represent full online interactions (Spence et al.,
2016). However, by doing this, the current study could focus on examining bandwagon
machine heuristics which was the central goal of the study.
The current study attempted to use a stranger’s Twitter account to prevent
implying any identity heuristics that might confound bandwagon heuristics. However, the
design used a male’s name “David” for the target user, which might suggest gender
differences. Thus, future research could use names for neutral gender to eliminate identity
cues. Another possibility would have been to simply use the default Twitter avatar for
this study. When individuals set up a Twitter account, and they do not upload an avatar or
profile picture, the default image is an egg-like object with a shadow. This may have
served as an appropriate way to structure the condition and should be considered for
future studies. However, this also can create problems because of participant’s
expectations about how an active Twitter page should be constructed. Participants might
perceive the lack of identity cues as indicating an inactive account and lead to other types of participant bias.
Consistently, regression analyses on demographics and Twitter daily usage found
that only gender played a significant role in predicting participants’ source competence
perceptions. Findings suggest that males tend to perceive that the target user as more
competent about the health risk topic in both conditions. Therefore, future studies should
investigate gender differences in source credibility perceptions within health and risk
context, specifically examining the role of social media.
69
Finally, although the results did not suggest a direct effect of the presence of retweet number on the judgments of source credibility, it is uncertain to what extent different retweet numbers would affect participants’ cognitive processing. Thus, future studies could keep examining the effects of various levels of retweet number.
Additionally, researchers in risk and crisis communication could consider how the number of retweets would serve in different risk and crisis forums. Moreover, the role of cognitive elaboration in risk message processing and cues has received some attention, it is still understudied. Future studies may examine what conditions or cues promote cognitive elaborations and what effects thinking more or less have on credibility judgments.
Study Two
Study Two uses an experimental design to examine how diverse author identity cues interacted with different levels of bandwagon heuristics on a health risk post, and effects on viewers’ source credibility perceptions. Nine experimental conditions within a health risk domain were compared. To be specific, participants viewed a Twitter profile page with the user identity of either an FDA expert, a peer, or a stranger posting the prevalence of contaminated watermelons in markets. Under the tweet about the watermelons, three different levels of retweet number (i.e., low – 40, moderate – 400, high – 4,000) were manipulated to elicit bandwagon heuristics. Participants responded to a measure of source credibility (McCroskey & Teven, 1999). The ratios used for the retweet number were based off previous research (e.g., Westerman et al., 2014;
Westerman et al., 2012).
70
The results did not indicate significant interaction effects between identity and bandwagon heuristics across three factors of source credibility, namely competence, goodwill, or trustworthiness. Rather, main effects for both cognitive heuristics were found on source credibility. The findings suggest that different online heuristic cues impact the judgments of competence, goodwill, and trustworthiness at different levels.
Theoretical and practical implications in terms of identity and bandwagon cues, as well as limitations and directions for future research, are discussed below.
Theoretical Implications
Identity cues may appear to be more robust than bandwagon cues in influencing source credibility in health and risk forum. Significant main effects for identity cues were found on all three factors of source credibility; whereas there was no significant main effect of retweet numbers on goodwill perceptions. In other words, individuals evaluate source credibility differently when viewing different Twitter user profile pages. Despite the levels of retweet number presented under the health risk tweet, individuals perceived that the Twitter user was putting the interest of the public in mind across all conditions.
More importantly, authority identity triggered by the FDA expert on Twitter was consistently perceived to be the most credible across three ANOVAs, which was significantly different from the peer and stranger identities. The findings harken back to health and risk literature on information seeking. People are prone to seek health and risk information directly from official government websites or the accounts of health officials, as these identities are more likely to provide high warrants for competence, trustworthiness, and goodwill (Sundar et al., 2009; Walther & Parks, 2002). Therefore, if
71 people are more likely to seek information from an expert, they may be more likely to rate an expert as credible.
Also, no significant differences were found between the conditions of peer and stranger user identities in inducing source credibility evaluations. Individuals perceived similar lower levels of source credibility when viewing tweets posted by a peer and a stranger, compared to viewing those from an expert. The results also indirectly support that individuals are more likely to follow authorities when a crisis emerges (Spence et al,
2005). Participants perceived low credibility in the peer and stranger conditions than those of an expert. Although the issue of contaminated watermelons is a highly relevant food safety concern for the participants, people may seek information sources with high expertise and technological abilities to provide instructions. Especially at the initial stage of a food safety crisis, people may question the competence and trustworthiness of a layperson or information.
Interestingly, participants evaluated the peer Twitter user as having higher competence and trustworthiness but lower goodwill than a stranger target user. This could be because once a poor perception is made about a sources competence and trustworthiness; this prevents one from judging specifically about their intentions. This explanation is speculative and requires further research.
For bandwagon heuristics, the results suggest that a larger bandwagon effect leads to low source credibility judgments, which, at first, appears counter-intuitive. In particular, regardless of the different identity heuristics represented by Twitter user profiles, participants perceived the lowest level of source credibility when viewing the risk tweet with the high number of retweets; especially for competence and
72 trustworthiness, which was significantly different from the tweets with low and moderate levels of retweets. Thus, one explanation is that it is possible that bandwagon heuristics, as predicted by the MAIN model, were not actually cued in this context. The findings may imply a resistance effect for high bandwagon heuristics in processing risk and health messages. A large retweet number may represent a point at which resistance has a stronger effect than the social influences of the bandwagon heuristics. Research on adoption of innovation within public health campaigns suggests that people tend to be more reluctant to extinguish an old behavior than to commence a new behavior (Rogers,
1983; Snyder et al., 2004). The manipulated tweet about contaminated watermelons suggest audiences to some extent, stop having watermelons to prevent sickness, which may be an undesirable adoption, or they are not ready to change their habitual behaviors
(Prochaska & DiClemente, 1983). Such resistance may develop with the increase of bandwagon heuristics, which may lead to negative impacts on source credibility evaluations. As participants’ willingness to change that specific behavior was not measured in the current study, future research should address this to determine if resistance impacts motivations.
Consistently, empirical studies on health campaigns have found resistance effects over time, such that the more and more people have advocated for a target health behavior, the remaining population became more resistant (Coleman, Katz, & Mendel,
1966; Hamblin, Jacobsen, & Miller, 1973). The resistance may be due to participants’ involvement with their behavioral change, their accumulative information or prior knowledge on the topic; or in this scenario their reactance against pressure exerted by the high-level retweet numbers (Hamilton & Stewart, 1993; Snyder et al., 2004). In this case,
73 the more retweets number indicated under the tweet, the more people that participants assume have engaged in the topic. If individuals respond with resistance to increased pressure to change, to higher levels of bandwagon heuristics, the greater the reactance.
Such reactance triggered by high bandwagon heuristics, in turn, undervalued the target user on source credibility.
The reverse bandwagon effect (Granovetter & Soong, 1986) or “snob” effects
(Leibenstein, 1950) induced by a large retweet number also provides an alternative explanation for the results, such that the more retweets of a health risk post, the lower its quality would be judged. Similar as resistance effects discussed above, bandwagon effects might not occur because of the valence of the risk topic associated with negative externalities (Hellosfs & Jacobson, 1999). Innovative behaviors could be viewed toward a positive valence; widespread retweets may lead to rallies and positive judgments of the content quality. However, health and risk topics are inevitably associated with a negative valence that has to be worked for and negotiated. Scholars in Economic Psychology, as well as research in Social Judgment Theory (Sherif & Hovland, 1961), argue that individuals who are not experts would primarily rely on their own beliefs (Lemert, 1986;
Pronin, Puccio, & Ross, 2002). Their preferable beliefs may serve as an anchor, making people more likely to underestimate the degree of heterogeneity in society (West, 1996).
As a result, individuals may tend to over-optimistically predict that the public has the similar opinions as they have (Babad, 1995; Bischoff & Egbert, 2013; Lemert, 1986).
In the contexts of risk and crises with controversies and uncertainties, such anchoring biases may lead to suspicions of the source and content when the risk messages are contradicted with wishful preferences. Thus, participants might perceive low
74 evaluations on source credibility with the widespread retweeting of the original post.
Future research on risk, crisis and health should examine the reverse bandwagon effects, to offer theoretical value for individuals’ learning process.
Moreover, the results for retweet numbers found no significant differences between the low and moderate levels of retweet number in inducing source credibility evaluations. The findings suggest that reverse bandwagon effects might not occur when the post was retweeted at a lower rate. It echoes with the results in Study One, where participants perceived moderate levels of source credibility regardless the presence of small retweet number.
It is also noticeable that participants perceived the highest levels of competence and goodwill when viewing the conditions with low retweet numbers (i.e., 40), which also consistent with the reverse bandwagon effects or resistance effects discussed above.
However, the results found the highest level of trustworthiness in the conditions with moderate retweet numbers (i.e., 400). As mentioned earlier, trustworthiness refers to the beliefs whether the target user would tell the truth as he or she knows it (O’Keefe, 1990), which to some degree may correlate with information circulation. Therefore, participants might perceive more trustworthiness about the target user when viewing the risk post with a moderate number of retweets than viewing that with an only low number of retweets.
Furthermore, descriptive statistic comparisons among all conditions found the highest levels of source credibility evaluations in the expert conditions with low or moderate retweets (i.e., 40 or 400). The results imply a cue-cumulative effect (Sundar,
Knobloch-Westerwick, & Hastall, 2007) which proposed that the co-presence of more
75
cues would enhance the positive effects of source credibility perceptions. In this case, the expert posted a tweet with few people retweeting, which might provide more identity
cues for credibility evaluations. The authority heuristics and reverse bandwagons work
together, making participants more susceptible to believe the message. Moreover, Sundar
et al. (2007) proposed that the cue-cumulative effect was invoked when all the cues only
trigger the same heuristics. The results further suggest cue-cumulative effect from both
bandwagon and identity heuristics.
Consistently, participants perceived the lowest levels of competence, goodwill, and trustworthiness in the stranger conditions with large retweets (i.e., 4,000). The results
also support the cue-cumulative effect. Bandwagon and peer identities boosted
participants’ credibility evaluations, compared to those bandwagon cues alone in the
stranger’s conditions. The results offer a novel consideration of the issues of user
identity, bandwagon cues, and perceived credibility in the online risk and health forum.
In general, several theoretical interpretations of the results together construct a view of
how different agency cues may interact during online risk information processing.
Practical Implications
As risk debates and crisis situations are featured with new evolving power
structures and substantial values, source credibility becomes a major medium for power
controls and social influence in such contexts (Renn & Levine, 1991). Despite the
advances empowered by social media technologies, social media also alter the public’s
information processing tendencies, thus creating challenges for locating credible and
trustworthy information, especially during extreme events with high threats and
76
uncertainties (Edwards et al., 2013). The findings regarding judgments of online source
credibility lead to several critical implications for public risk and health practitioners.
First, given the consistent findings in user identity with previous research, it is
necessary to employ authority cues to communicate with the public and organizational stakeholders. Twitter accounts with experts’ names may serve as official identity cues to guide individuals’ credibility perceptions. Empirical cases such as the CDC have
evidenced the application of authority cues in online health communication. Government
agencies and public health organizations at the federal, state and local levels should be
encouraged to set up official social media accounts with official identities on the profiles
in order to convey credible perceptions when communicating with the public or partners.
Second, given the cue-cumulative effects, public health practitioners and
emergency managers should be encouraged to cooperate with the public, similar
organizations, and stakeholders from the targeted communities in online conversations. In
accordance with Seeger (2006), in order “to maintain effective networks, crisis planners
and communicators should continuously seek to validate sources, choose subject-area experts, and develop relationships with stakeholders at all levels” (p. 240). Social media provides both an opportunity and mechanism for members of the public to participate in the crisis discussion. Besides professional voices, emergency managers should invite peers or stakeholders from the targeted communities, to build the trustworthy and healthy relationships with the public, which may generate a cue-cumulative effect.
Third, given the resistant or reverse bandwagon effects on source credibility perceptions, public health practitioners need to bear in mind that members of the public who are sharing information, don’t have to take an expressed mission in an extreme
77 event. Information overload might weaken credible voices especially given visual anonymities online, which may raises suspicions and rumors around a risk topic. Thus, emergency managers and crisis practitioners must not only monitor the social media landscape, but also be cautious about bandwagon affordances, such as retweet numbers.
For instance, to reduce the potential of reverse bandwagon effects triggered by retweet numbers, emergency practitioners could also update the tweets frequently. By doing this, it may be possible to reduce the high ratio of retweet number displayed on each original post.
Limitation and Future Directions
Limitations from Study Two are discussed next. First, the experiment used static
Twitter page for each condition which may decrease ecological validity. Similar to Study
One and previous social media affordances studies, captured a screenshot of the mock
Twitter pages were presented to ensure the internal validity of the experiment. Such designs ran short in providing the “live running” nature of an actual Twitter feed, which could not represent full online interactions (Spence et al., 2016). Participants may have different perceptions in source credibility evaluations if they could review other bandwagon and identity heuristics such as comments for the risk health tweet. However, by doing this, the current study could not focus on examining bandwagon machine heuristics.
The results are based on self-report data from a convenient sample. As the majority of the participants in the current study were younger Caucasian adults with high economic status from the United States of America, the results may fall short in predicting for the overall population, especially some underserved minority groups who
78
often are disproportionately at risk. Therefore, future research could employ more
representative samples from the population. Also, the current study uses experimental
design and focuses more on theoretical examinations. Empirical studies should be conducted with field experiments to observe how various agency cues work in a real context to impact credibility perceptions.
Importantly, three levels of retweet numbers represent different ratios of bandwagon heuristics in the research design, revealing reverse bandwagon effects on participants source credibility perceptions. For empirical health and risk communication, it is important to know when the bandwagon cues start to show a reverse effect.
Additionally, the current study examines identity cues at the individual level. The findings provide theoretical evidence that individual experts are perceived to be the most credible in a risk and health topic; future research could examine the extent of which organizational authorities impact people’s credibility perceptions.
Furthermore, although there were no significant interaction effects between the user identity and retweet number in eliciting source credibility judgments, significant interactions effects were evidenced in other studies (e.g., Lee & Sundar, 2013). Thus, scholars could further examine the bandwagon heuristics within different contexts to offer empirical implications and further research.
Study Three
Study Three uses an experimental design to examine bandwagon effects on credibility judgments when the information source is at the organizational level. Three experimental conditions within a health risk domain were compared. Consistent with the
79
previous two studies, participants viewed a health risk tweet on a Twitter profile page
about the presence of contaminated watermelons in markets. Rather than using multiple
individual levels of user identities across conditions, the current study focuses on
authority identity, and used the FDA official Twitter page to trigger organizational
identity heuristics. Consistent with the design of Study Two, under the tweet about the
risk, three different levels of retweet numbers (i.e., low – 40, moderate – 400, high –
4,000) were manipulated to elicit bandwagon heuristics. Participants then responded to
RAND Public Health Trust Scale (Eisenman et al., 2012). Similar to Study Two, the ratios used for the retweet number were based off previous research (e.g., Westerman et al., 2014; Spence et al., 2016).
The results found significant differences for credibility judgments on the organization among the three conditions. In accordance with Study Two, a consistent reverse bandwagon effect was observed when participants viewed different levels of retweet numbers, which contributes theoretical value to the MAIN model. Theoretical and practical implications in terms of heuristic cues, as well as limitations and directions for future research, are discussed below.
Theoretical Implications
Bandwagon heuristics hold a similar pattern in influencing credibility judgments when information sources are at the organizational level. The results reveal that the higher levels of retweet numbers presented in the FDA profile page, the lower level of credibility evaluations participants made about the FDA. In accordance with Study Two, upward trends of bandwagon heuristics, as predicted by the MAIN model, were not actually triggered in the current context. The findings in both studies add theoretical
80 value to the MAIN model, as social media heuristics may trigger reverse bandwagon effects for the cognitive process. The findings provide evidence that bandwagon heuristics are consistent when user identities, especially authority identities, are presented on both organizational and individual levels, which would negatively induce credibility perceptions.
It is noticeable that the high level of retweets (i.e., 4,000) lead to the lowest levels of credibility perceptions in the current study and Study Two. The findings suggest a reverse bandwagon effect in credibility perceptions on organizations, which may also imply a resistance effect for high bandwagon heuristics in processing risk and health messages. Harkening back to the interpretations in Study Two, a large size of populations involved in a health risk topic may increase an individual’s reactance to adoptions. Such resistance may be due to individuals’ psychological discomforts of withdrawal, readiness to change (Prochaska & DiClemente, 1983), accumulated information on the topic
(Hamilton & Stewart, 1993), interpersonal influences, or their life circumstance (Hovland
& Janis, 1953). Therefore, the more that people have engaged in the behavior or the topic, the more the likelihood that the viewer might respond in a maladaptive manner, which may, in turn, lead to a lower credibility judgment on the source.
Moreover, given the nonreciprocal relationships available on social media, the retweet number may also present a perceived anonymity of topic advocators. Previous studies on computer-mediated communication (CMC) indicate that the perceptions of anonymity affect both pro- and anti-social behavior performances (Joinson, 2001). In line with deindividualization caused by online anonymity (Spears & Lea, 1994), a state of decrease in identity awareness may lead to a decline of adherences to societal norms,
81 which, in turn, decreases the social influences of the bandwagon effects. Although the
FDA posted the original tweet, the potential anonymity afforded by retweet number implies identity uncertainties. As a result, because a large number of retweets was presented for this health risk tweet, it is likely that participants perceived online users as not identifiable, which may raise their suspicions on the content quality of the messages, such as credibility perceptions about the original tweet.
Also, participants perceived no significant differences in the credibility of the
FDA when viewing conditions with low (i.e., 40) and moderate (i.e., 400) levels of retweets. The findings are consistent with both Study One and Study Two. A trend of reverse bandwagon effects might not occur when the health risk post was retweeted at a lower rate. However, the participants in Study Two perceived the highest level of credibility evaluations when viewing moderate level of retweets with the individual authority cues (i.e., FDA expert with 400 retweets); whereas the participants in the current study perceived the highest level of credibility in the condition of low retweet level with organizational authority cues (i.e., FDA organization with 40 retweets).
Therefore, it is possible that the reverse bandwagon effects occur earlier when author identities are organizations rather than individuals.
One possibility for the interpretations may attribute certain levels of bandwagon effects to the organizational user identity. When viewing the FDA’s official page, participants may also perceive an organization as the sum of individual experts, which may imply bandwagon heuristics. Perceptions of an organizational user identity may add values to bandwagon heuristics. In this case, people exposed to the profile page of the
FDA official account may perceive the accumulative identities compared to those
82 viewing a profile page from certain FDA expert. The organizational user identity may imply accumulative identities, which may suggest both authority and bandwagon heuristics. Consistent with the MAIN model, the findings imply that the same media affordances may trigger different cognitive heuristics at the same time.
Additionally, the current study applied organizational level measurements to evaluate participants’ perceptions of an organization. Although the means of the credibility perceptions were significantly different across the three conditions, the actual scores were all above the scale midpoint. This might be because the FDA has been recognized with high credibility and reputations in health and risk management.
Consistent with previous studies, people may perceive higher credibility for the official agencies such as CDC and FDA than state agencies and private sectors (Spence et al.,
2016). It is possible that the pre-existed organizational reputations might boost the degree of confidence in the self-report data. Therefore, future research may use pre- and post- tests to evaluate participants cognitive outcomes on the experimental stimuli. Moreover, the results are consistent with previous studies, such as Westerman et al., (2014), that used individual-level scales for cognitive measurements on public health agencies. Future studies should examine if there were an outcome consistency when using other organizational measurements on organization variables on other identities.
Practical Implications
Risk and crisis literature, as well as research in public health and epidemiology, has provided numerous suggestions concerning how to best meet the public’s needs of crisis and risk responses, how to address stakeholders’ emotional concerns, and many other aspects of emergency management (Reynold, 2006; Veil, Reynold, Sellnow, &
83
Seeger, 2008). However, recommendations for the use of social media for crisis
management are largely unstructured and untested (Lachlan et al., 2014; Spence, Lachlan
& Rainear, 2016). The current study provides experimental evidence to suggest several practical implications, which contributes to best practice of risk and crisis communication in social media.
First, the results across the three studies consistently suggest a reverse bandwagon effect for the displays of retweet numbers on credibility judgments. Therefore, instead of passively disseminating risk and crisis information online, emergency managers, as well as public health practitioners, should fully adopt social media technologies for two-way
crisis communication, to monitor online activities to establish situational awareness
among the public.
It is evidenced that government agencies tend to provide a one-way stream of information to their stakeholders (Waters & Williams, 2011). In this case, such one-way communication was reflected in the bandwidth of retweets that suggests a non-reciprocal, unidirectional interactions between message senders and receivers. The current findings echo with empirical evidence. At one point during 2012 Hurricane Sandy, there were over 12.5 tweets sent per second using the hashtags promoted by NOAA and FEMA; however, in a sample of tweets collected cross the storm, only nine tweets directly came from government agencies (Lachlan et al., 2014a; 2014b). During the information glut of everyone commenting on an emergency, government agencies and crisis managers should actively engage in the dialogue, seek the attention of affected publics, and try to make sense of an overflow of information.
84
Second, given that reverse bandwagon effects may be more likely to occur from organizational user identities than from individual user identities, public health practitioners and emergency managers should carefully plan and monitor the schedules and frequencies of social media feeds. For instance, official accounts could send out risk and crisis information via a series of tweets over and over again rather than a single tweet through the duration of an extreme event, which may be helpful to reduce the appearance of retweet amounts for a single post. Effective crisis management should focus on providing constant information feeds on social media, to reduce the bandwidth of retweets from one single post. By doing so, the public would engender a sense that government agencies and organizations have the best interest of the audience in mind
(Lachlan et al., 2014a: Spence et al., 2016).
Third, organizational-level identities may trigger both identity and bandwagon heuristics. Regarding the potential resistance effects of bandwagon heuristics, agencies and organizations should also encourage their members to set up individual official accounts, to interact with both the official social media feeds and the public. By doing this, public health agencies and emergency managers could build an alignment of authority heuristics on different levels, which may promote cue-cumulative effects on message credibility evaluations and other online information processing.
Limitations and Future Directions
Considering the consistency of experimental designs with the previous two studies; Study Three bears similar limitations from the design features. For instance, the current study used a static Twitter page for each condition, which may decrease ecological validity. Participants were recruited from a convenient sample with relatively
85
high homogeneity regarding basic demographic characteristics. Additional research could
examine how retweet numbers or other bandwagon cues displayed by official agencies
would affect the public’s cognitive perceptions in real-time food safety events.
As a pioneering trial to compare cognitive processing between individual- and
organizational-level sources within health and risk, Study Three provides experimental
evidence on how bandwagon heuristics would influence credibility judgments when
interacting with an organizational-level authority identity. However, it is not clear to what extent the findings from this study would apply to situations when source identities do not contain authority cues. Future research could investigate how bandwagon heuristics impact risk and health cognitive processing when information sources are from the private sector or local communities.
Consistent with Study Two, the current study employed the same three levels of retweet numbers to represent different ratios of bandwagon heuristics. However, given that organizational identity cues may also trigger bandwagon heuristics, future research should investigate the tipping points for the reverse bandwagon effects, and compare the thresholds between individual and official sources.
The tweet in the current design may also reflect a food safety crisis at its early
stage. Therefore, scholars should investigate if participants perceived a constant reverse
bandwagon effect on the judgments of a public health agency’s credibility throughout the
event. In addition, future research could examine how online bandwagon heuristics
influence different cognitive outcomes about an organization.
86
Chapter Summary
This chapter provided the research discussions regarding the study results including theoretical and practical implications for the MAIN model and risk and crisis communication. Authority identity was suggested to be the most stable indicator for crediblity evaluations; a resistant or reverse bandwagon effect was discussed. Finally, each study elaborated the research limitations and future directions.
Copyright © Xialing Lin 2016
87
CHAPTER 6: CONCLUSION
Drawn from the theories of online cognitive processing, this research suggests
that bandwagon and identity heuristics triggered by social media affordances, including
retweet numbers and user identities on Twitter, may influence judgments of credibility in
a health risk situation. The theoretical and practical implications of these results are wide-ranging. First, the current study extends a long line of work on the heuristics cues on online cognitive processing and adds values to the current media effect theories such as the MAIN model. Second, the findings provide experimental evidence for risk and crisis literature that online media affordances may influence credibility perceptions on a health risk topic.
To be specific, the hypothesis proposed by Study One did not receive significant support from the data. Bandwagon heuristics triggered by a number of retweets did not influence people’s source credibility. However, a mediation effect by cognitive elaboration was found in the path analysis of the examined variables, which suggest indirect predictions for bandwagon heuristics. Study one also suggest a gender difference in perceiving source credibility on the given health risk topic which should be examined in future studies.
Study Two intends to examine the interactions between bandwagon and identity heuristics on social media. Moreover, Study Three aims to compare the effects of bandwagon heuristics on credibility perceptions between organizational and individual level author identities. Authority cues appear to be a vital factor in influencing cognitive perceptions compared to peer and stranger identities. Importantly, a resistant effect or a reverse bandwagon effect was found in both studies, which provide theoretical and
88 practical insights to risk and crisis communication. Overall, examining the online media affordances to influence risk and crisis cognitive processing provides further insights into both online media effects and crisis and risk management.
Copyright © Xialing Lin 2016
89
Appendix A
Manipulation Example for Study One
90
Appendix B
Authority Manipulation Example for Study Two
91
Appendix C
Peer Manipulation Example for Study Two
92
Appendix D
Stranger Manipulation Example for Study Two
93
Appendix E
Manipulation Example for Study Three
94
References
Babad, E. (1995). Can accurate knowledge reduce wishful thinking in voters' predictions
of election outcomes? The Journal of Psychology, 129(3), 285-300.
doi:10.1080/00223980.1995.9914966
Bean, H., Liu, B. F., Madden, S., Sutton, J., Wood, M. M., & Mileti, D. S. (2016).
Disaster warnings in your pocket: How audiences interpret mobile alerts for an
unfamiliar hazard. Journal of Contingencies and Crisis Management.
doi:10.1111/1468-5973.12108
Berlo, D. K., Lemert, J. B., & Mertz, R. J. (1969). Dimensions for evaluating the
acceptability of message sources. Public Opinion Quarterly, 33(4), 563-576.
doi:10.1086/267745
Bischoff, I., & Egbert, H. (2013). Social information and bandwagon behavior in voting:
An economic experiment. Journal of Economic Psychology, 34, 270-284.
doi:http://dx.doi.org/10.1016/j.joep.2012.10.009
Bollen, K. A. (1989). Structural equations with latent variables. New York, NY: John
Wiley.
Bunz, M. (January 14th, 2010). In Haiti earthquake coverage, social media gives victim a
voice. Retrieved March 25th, 2013 from
http://www.guardian.co.uk/media/pda/2010/jan/14/socialnetworking-haiti
Campbell, D. T., & Stanley, J. C. (1963). Experimental and experimental designs for
research. Chicago, IL: Rand-McNally.
Cappelli, P., & Sherer, P. D. (1991). The missing role of context in OB-the need for a
meso-level approach. Research in Organizational Behavior, 13, 55-110.
95
Chen, S., & Chaiken, S. (1999). The heuristic-systematic model in its broader context. In
S. Chaiken & Y. Trope (Eds.), Dual-process theories in social psychology (pp.
73-96). New York, NY: Guilford.
Cheong, M., & Lee, V. S. (2011). A microblogging-based approach to terrorism
informatics: Exploration and chronicling civilian sentiment and response to
terrorism events via Twitter. Information Systems Frontiers, 13(1), 45-59.
doi:10.1007/s10796-010-9273-x
Chevalier, J. A., & Mayzlin, D. (2006). The effect of word of mouth on sales: Online
book reviews. Journal of Marketing Research, 43(3), 345-354.
doi:doi:10.1509/jmkr.43.3.345
Coleman, J.S., Katz, E., & Menzel, H. (1966). Medical innovation: A diffusion study.
Indianapolis: Bobbs-Merrill.
Cronkhite, G., & Liska, J. (1976). A critique of factor analytic approaches to the study of
credibility. Communication Monographs, 43(2), 91-107.
doi:10.1080/03637757609375920
Csikszentmihalyi, M., & Kubey, R. (1981). Television and the rest of life: A systematic
comparison of subjective experience. Public Opinion Quarterly, 45(3), 317–28.
DeAndrea, D. C. (2014). Advancing warranting theory. Communication Theory, 24(2),
186-204. doi:10.1111/comt.12033
Donath, J. (1999). Identity and deception in the virtual community. In M. A. Smith & P.
Kollock (Eds.), Communities in cyberspace (pp. 29-59). New York, NY:
Routledge.
96
Donath, J. (2007). Signals in social supernets. Journal of Computer-Mediated
Communication, 13, 231–251. doi: 10.1111/j.1083-6101.2007.00394.x
Eastin, M. S. (2006). Video game violence and the female game player: Self- and
opponent gender effects on presence and aggressive thoughts. Human
Communication Research, 32(3), 351-372. doi:10.1111/j.1468-
2958.2006.00279.x
Edwards, C., Edwards, A., Spence, P. R., & Shelton, A. K. (2014). Is that a bot running
the social media feed? Testing the differences in perceptions of communication
quality for a human agent and a bot agent on twitter. Computers in Human
Behavior, 33, 372-376. doi: 10.1016/j.chb.2013.08.013
Eisenman, D.P., Williams, M.V., Glik, D., Long, A., Plough, A.L., & Ong, M. (2012).
The public health disaster trust scale: Validation of a brief measure. Journal of
Public Health Management and Practice, 18(4), E11-E18. doi:
10.1097/PHH.0b013e31823991e8
Flanagin, A. J., & Metzger,M. J. (2013). Trusting expert- versus user-generated ratings
online: The role of information volume, valence, and consumer characteristics.
Computers in Human Behavior, 29, 1626–1634. doi: 10.1016/j.chb.2013.02.001
Fogg, B. J., & Tseng, H. (1999). The elements of computer credibility. Paper presented at
the Proceedings of the SIGCHI conference on Human Factors in Computing
Systems, Pittsburgh, Pennsylvania, USA.
Fox, S. (2011, May 12). The social life of health information, 2011. Pew Research
Center's Internet & American Life Project. Washington, D.C. Retrieved from
97
http://www.pewinternet.org/˜/media//Files/Reports/2011/PIP_Social_Life_of_Hea
lth_Info.pdf
Freberg, K., & Palenchar, M. J. (2013). Convergence of digital negotiation and risk
challenges: strategic implications of social media for risk and crisis
communications. In H. S. N. Al-Deen and J. A. Hendricks (Eds.), Social media
and strategic communications (pp. 83-100). New York, NY: Palgrave Macmillan.
Gaziano, C., & McGrath, K. (1986). Measuring the concept of credibility. Journalism &
Mass Communication Quarterly, 63(3), 451-462.
doi:10.1177/107769908606300301
Glik, D. C. (2007). Risk communication for public health emergencies. Annual Review of
Public Health, 28(1), 33-54. doi:10.1146/annurev.publhealth.28.021406.144123
Goolsby, R. (2010). Social media as crisis platform: The future of community maps/crisis
maps. ACM Transactions on Intelligent Systems and Technology, 1(1), 1-11.
doi:10.1145/1858948.1858955
Granovetter, M., & Soong, R. (1986). Threshold models of interpersonal effects in
consumer demand. Journal of Economic Behavior & Organization, 7(1), 83-99.
doi: 10.1016/0167-2681(86)90023-5
Hackman, J. R. (2003). Learning more by crossing levels: Evidence from airplanes,
hospitals, and orchestras. Journal of Organizational Behavior, 24(8), 905-922.
doi:10.1002/job.226
Hamblin, R.L., Jacobsen, R.B., & Miller, J.L.L. (1973). A mathematical theory of social
change. New York: Wiley.
98
Hamilton, M.A. & Stewart, B. L. (1993). Extending an information processing model of
language intensity effects. Communication Quarterly, 41, 231–246.
Hellofs, L. L., & Jacobson, R. (1999). Market share and customers' perceptions of
quality: When can firms grow their way to higher versus lower quality? Journal
of Marketing, 63(1), 16-25. doi:10.2307/1251998
Hilligoss, B., & Rieh, S. Y. (2008). Developing a unifying framework of credibility
assessment: Construct, heuristics, and interaction in context. Information
Processing & Management, 44(4), 1467-1484. doi:10.1016/j.ipm.2007.10.001
Hu, Y., & Sundar, S. S. (2009). Effects of online health sources on credibility and
behavioral intentions. Communication Research. doi:10.1177/0093650209351512
Hurd, P. L., & Enquist, M. (2005). A strategic taxonomy of biological communication.
Animal Behavior, 70, 1155–1170. doi:10.1016/j.anbehav.2005.02.014
Hovland, C. I., Janis, I. L., & Kelley, H. H. (1953). Communication and persuasion:
Psychological studies of opinion change. New Haven, CT: Yale University Press.
Joinson, A. N. (2001). Self-disclosure in computer-mediated communication: The role of
self-awareness and visual anonymity. European Journal of Social Psychology,
31(2), 177-192. doi:10.1002/ejsp.36
Kaplan, A. M., & Haenlein, M. (2010). Users of the world, unite! The challenges and
opportunities of Social Media. Business horizons, 53(1), 59-68.
Kim, K. J., Park, E., & Sundar, S. S. (2012). Ips vs. Amoled: Effects of panel type on
smartphone users’ viewing and reading experience. In H. J. J. Park, Q. Jin, M.
Sang-soo Yeo, & B. Hu (Eds.), Human centric technology and service in smart
space: Humancom 2012 (pp. 77-84). Dordrecht: Springer Netherlands.
99
Kline, R. B. (1998). Principles and practice of structural equation modeling: An
introduction to LISREL and EQS. New York, NY: Springer.
Lachlan, K. A., Spence, P. R., Edwards, A., Reno, K. M., & Edwards, C. (2014a). If you
are quick enough, I will think about it: Information speed and trust in public
health organizations. Computers in Human Behavior, 33, 377-380.
doi:http://dx.doi.org/10.1016/j.chb.2013.08.014
Lachlan, K.A., Spence, P.R., & Lin, X. (2014b). Expressions of risk awareness and
concern through Twitter: On the utility of using the medium as an indication of
audience needs. Computers in Human Behavior, 35, 554-559. doi:
http://dx.doi.org/10.1016/j.chb.2014.02.029
Lachlan, K. A., Spence, P. R., Lin, X., Najarian, K. M., & Greco, M. D. (2014c). Twitter
use during a weather event: Comparing content associated with localized and
nonlocalized hashtags. Communication Studies, 65(5), 519-534.
doi:10.1080/10510974.2014.956940
Lachlan, K. A., Spence, P. R., Lin, X., Najarian, K., & Del Greco, M. (2015). Social
media and crisis management: CERC, search strategies, and twitter content.
Computers in Human Behavior. doi:http://dx.doi.org/10.1016/j.chb.2015.05.027
Lachlan, K. A., Westerman, D. K., & Spence, P. R. (2010). Disaster news and subsequent
information seeking: Exploring the role of spatial presence and perceptual
realism. Electronic News, 4(4), 203-217. doi:10.1177/1931243110387092
Lee, J. Y., & Sundar, S. S. (2013). To tweet or to retweet? That is the question for health
professionals on twitter. Health Commun, 28(5), 509-524.
doi:10.1080/10410236.2012.700391
100
Leibenstein, H. (1950). Bandwagon, snob, and Veblen effects in the theory of consumers'
demand. The Quarterly Journal of Economics, 64(2), 183-207.
doi:10.2307/1882692
Lemert, J. B. (1986). Picking the winners: Politician vs. Voter predictions of two
controversial ballot measures. The Public Opinion Quarterly, 50(2), 208-221.
Retrieved from http://www.jstor.org/stable/2748884
Levine, T. R. (2005). Confirmatory factor analysis and scale validation in communication
research. Communication Research Reports, 22(4), 335-338.
doi:10.1080/00036810500317730
Levine, T., Hullett, C. R., Turner, M. M., & Lapinski, M. K. (2006). The desirability of
using confirmatory factor analysis on published scales. Communication Research
Reports, 23(4), 309-314. doi:10.1080/08824090600962698
Li, R., Zhang, B., Sundar, S. S., & Duh, H. B.-L. (2013). Interacting with augmented
reality: How does location-based ar enhance learning? In P. Kotzé, G. Marsden,
G. Lindgaard, J. Wesson, & M. Winckler (Eds.), Human-computer interaction –
interact 2013: 14th ifip tc 13 international conference, cape town, south africa,
september 2-6, 2013, proceedings, part ii (pp. 616-623). Berlin, Heidelberg:
Springer Berlin Heidelberg.
Lim, Y., & Van Der Heide, B. (2015). Evaluating the wisdom of strangers: The perceived
credibility of online consumer reviews on yelp. Journal of Computer-Mediated
Communication, 20(1), 67-82. doi:10.1111/jcc4.12093
Limperos, A. M., Buckner, M. M., Kaufmann, R., & Frisby, B. N. (2015). Online
teaching and technological affordances: An experimental investigation into the
101
impact of modality and clarity on perceived and actual learning. Computers &
Education, 83, 1-9. doi:10.1016/j.compedu.2014.12.015
Madden, M., & Fox, S. (2006). Riding the waves of ‘Web 2.0’. Pew internet and
American life project, 5. Retrieved from http://clearwww.co-
bw.com/Ethics/ethics_Web_2.pdf
McCroskey, J. C., & Teven, J. J. (1999).Goodwill: A reexamination of the construct and
its measurement. Communication Monographs, 66, 90-103. doi:
10.1080/03637759909376464
McCroskey, J. C., & Young, T. J. (1981). Ethos and credibility: The construct and its
measurement after three decades. Central States Speech Journal, 32(1), 24-34.
doi:10.1080/10510978109368075
Metzger, M. J., & Flanagin, A. J. (2015). Psychological approaches to credibility
assessment online. In S. S. Sundar (Ed.), The handbook of the psychology of
communication technology (pp. 445-466). Hoboken, NJ: Wiley-Blackwell.
Metzger, M. J., Flanagin, A. J., Eyal, K., Lemus, D. R., & McCann, R. M. (2003).
Credibility for the 21st century: Integrating perspectives on source, message, and
media credibility in the contemporary media environment. Communication
yearbook, 27, 293-336
Metzger, M.J., Flanagin, A.J., & Medders, R.B. (2010). Social and heuristic approaches
to credibility evaluation online. Journal of Communication, 60(3), 413-439. doi:
10.1111/j.1460-2466.2010.01488.x
Nimmo, D., & Savage, R. L. (1976). Candidates and their images: Concepts. methods
and findings. Pacific Palisades: Goodyear.
102
O'Keefe, D. J. (1990). Persuasion: Theory and Research. Thousand Oaks: CA: Sage.
Omilion-Hodges, L. M., & Rodriguez, M. (2014). The effect of system generated cues on
users perceptions of organizational trustworthiness, competence and goodwill.
Journal of Media and Communication Studies, 6(10), 161-164.
Priester, J. R., & Petty, R. E. (1995). Source attributions and persuasion: Perceived
honesty as a determinant of message scrutiny. Personality and Social Psychology
Bulletin, 21(6), 637-654. doi:10.1177/0146167295216010
Prochaska, J. O. & DiClemente, C. C. (1983). Stages and processes of self-change of
smoking: Toward and integrative model of change. Journal of Consulting and
Clinical Psychology, 51, 390–395.
Pronin, E., Puccio, C., & Ross, L. (2002). Understanding misunderstanding: Social
psychological perspectives. In T. Gilovich, D. Griffin, & D. Kahneman (Eds.),
Heuristics and biases (pp. 636 - 665). Cambridge: Cambridge University Press.
doi: 10.1017/CBO9780511808098.038
Pure, R. A., Markov, A. R., Mangus, J. M., Metzger, M. J., Flanagin, A. J., & Hartsell, E.
H. (2012). Understanding and evaluating source expertise in an evolving media
environment. In T. Takseva (Ed.), Social software and the evolution of user
expertise: future trends in knowledge creation and dissemination: Future trends
in knowledge creation and dissemination (pp. 37-51). Hershey, PA: Information
Science Reference.
Renn, O., & Levine, D. (1991). Credibility and trust in risk communication. In R. E.
Kasperson and P. J. M. Stallen (eds.), Communicating risks to the public. (pp.
175–218). Dordrecht, NL: Kluwer.
103
Reuter, C. (2015). Crisis 2.0: Towards a Systematization of Social Software Use
(IJISCRAM). In Emergent collaboration infrastructures (pp. 35-48). Springer
Fachmedien Wiesbaden.
Reynolds, B. (2006). Response to Best Practices. Journal of Applied Communication
Research, 34(3), 249-252. doi:10.1080/00909880600771593
Reynolds, B., & W. Seeger, M. (2005). Crisis and emergency risk communication as an
integrative model. Journal of Health Communication, 10(1), 43-55.
doi:10.1080/10810730590904571
Rieh, S. Y., & Danielson, D. R. (2007). Credibility: A multidisciplinary framework.
Annual Review of Information Science and Technology, 41(1), 307-364.
doi:10.1002/aris.2007.1440410114
Rogers, E. M. (1983). Diffusion of innovations (3rd Ed.). New York: Free Press.
Schneider, B., Goldstiein, H. W., & Smith, D. B. (1995). The asa framework: An update.
Personnel Psychology, 48(4), 747-773. doi:10.1111/j.1744-6570.1995.tb01780.x
Sellnow, T. L., Ulmer, R. R., Seeger, M. W., & Littlefield, R. S. (2009). Effective risk
communication: A message-centered approach. New York: Springer.
Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and experimental
designs for generalized causal inference. Belmont, CA: Wadsworth Cengage
learning.
Sherif, M., & Hovland, C. I. (1961). Social judgment: Assimilation and contrast effects in
communication and attitude change. Oxford, England: Yale University.
Smith, J. M., & Harper, D. G. C. (1995). Animal signals: Models and terminology.
Journal of Theoretical Biology, 177, 305–311. doi:10.1006/jtbi.1995.0248
104
Snyder, L. B., Hamilton, M. A., Mitchell, E. W., Kiwanuka-Tondo, J., Fleming-Milici,
F., & Proctor, D. (2004). A meta-analysis of the effect of mediated health
communication campaigns on behavior change in the United States. Journal of
Health Communication, 9, 71-96. doi:10.1080/10810730490271548
Spears, R., & Lea, M. (1994). Panacea or panopticon? The hidden power in computer-
mediated communication. Communication Research, 21(4), 427-459.
doi:10.1177/009365094021004001
Spence, M. (2002). Signaling in retrospect and the informational structure of markets.
The American Economic Review, 92, 434–459. doi:10.1257/00028280260136200
Spence, P. R., Lachlan, K. A., Edwards, A., & Edwards, C. (2016). Tweeting fast
matters, but only if I think about it: Information updates on social media.
Communication Quarterly, 64(1), 55-71. doi:10.1080/01463373.2015.1100644
Spence, P.R., Lachlan, K.A., Lin, X., & del Greco, M. (2015a). Variability in Twitter
content across the stages of a natural disaster: Implications for crisis
communication. Communication Quarterly, 63(2), 171-186. doi:
10.1080/01463373.2015.1012219
Spence, P. R., Lachlan, K. A., Lin, X., Sellnow-Richmond, D. D., & Sellnow, T. L.
(2015b). The problem with remaining silent: Exemplification effects and public
image. Communication Studies, 1-17. doi:10.1080/10510974.2015.1018445
Spence, P. R., Lachlan, K. A., Burke, J. M., & Seeger, M. W. (2007). Media use and
information needs of the disabled during a natural disaster. Journal of Health
Care for the Poor and Underserved, 18, 394-404.
105
Spence, P. R., Lachlan, K. A., & Rainear, A. M. (2016). Social media and crisis research:
Data collection and directions. Computers in Human Behavior, 54, 667-672.
doi:http://dx.doi.org/10.1016/j.chb.2015.08.045
Spence, P. R., Sellnow-Richmond, D., Sellnow, T., & Lachlan, K. A. (in press). Social
media and corporate reputation during crises: The viability of video sharing
websites for providing counter-messages to traditional broadcast news. Journal of
Applied Communication Research.
Spence, P. R., Westerman, D., Skalski, P., Seeger, M., Ulmer, R. R., Venette, S., &
Sellnow, T. (2005). Proxemic effects on information seeking after the September
11th attacks. Communication Research Reports, 22, 39-46.
Starbird, K., Palen, L., Hughes, A., & Vieweg, S. (2010). Chatter on the red: What
hazards threat reveals about the social life of microblogged information. In Proc.
CSCW 2010.
Staw, B.M. & Sutton, R.I. (1992). Macro Organizational Psychology. In J.K.
Murnighan(ed.), Social psychology in organizations: Advances in theory and
research. Engiewood Cliffs, NJ: Prentice-Hall.
Sundar, S. S. (1998). Effect of source attribution on perception of online news stories.
Journalism & Mass Communication Quarterly, 75(1), 55-68.
doi:10.1177/107769909807500108
Sundar, S. S. (2008). The MAIN model: A heuristic approach to understanding
technology effects on credibility. In M. J. Metzger & A. J. Flanagin (Eds.),
Digital media, youth, and credibility (pp. 73–100). Cambridge, MA: The MIT
Press
106
Sundar, S. S., Jia, H., Waddell, T. F., & Huang, Y. (2015). Toward a theory of interactive
media effects (TIME): Four models for explaining how interface features affect
user psychology. In S. S. Sundar (Ed.), The handbook of the psychology of
communication technology (pp. 47-86). Malden, MA: Wiley Blackwell.
Sundar, S. S., & Kalyanaraman, S. (2004). Arousal, memory, and impression-formation
effects of animation speed in web advertising. Journal of Advertising, 33(1), 7-17.
doi:10.1080/00913367.2004.10639152
Sundar, S. S., Knobloch-Westerwick, S., & Hastall, M. R. (2007). News cues:
Information scent and cognitive heuristics. Journal of the American Society for
Information Science and Technology, 58(3), 366-378. doi:10.1002/asi.20511
Sundar S. S., & Nass, C. (2000). Source orientation in human-computer interaction:
Programmer, networker, or independent social actor. Communication Research,
27(6), 683-703. doi:10.1177/009365000027006001
Sundar, S. S., & Nass, C. (2001). Conceptualizing sources in online news. Journal of
Communication, 51(1), 52-72. doi:10.1111/j.1460-2466.2001.tb02872.x
Sundar, S. S., Oeldorf-Hirsch, A., & Garga, A. (2008). A cognitive-heuristics approach to
understanding presence in virtual environments. In PRESENCE 2008:
Proceedings of the 11th Annual International Workshop on Presence (pp. 219-
228).
Sundar, S. S., Xu, Q., & Oeldorf-Hirsch, A. (2009). Authority vs. Peer: How interface
cues influence users. Paper presented at the CHI '09 Extended Abstracts on
Human Factors in Computing Systems, Boston, MA, USA.
107
Sutton, J., League, C., Sellnow, T. L., & Sellnow, D. D. (2015). Terse messaging and
public health in the midst of natural disasters: The case of the boulder floods.
Health Commun, 30(2), 135-143. doi:10.1080/10410236.2014.974124
Tong, S. T., Van Der Heide, B., Langwell, L., & Walther, J. B. (2008). Too much of a
good thing? The relationship between number of friends and interpersonal
impressions on facebook. Journal of Computer-Mediated Communication, 13(3),
531-549. doi:10.1111/j.1083-6101.2008.00409.x
Utz, S. (2010). Show me your friends and i will tell you what type of person you are:
How one's profile, number of friends, and type of friends influence impression
formation on social network sites. Journal of Computer-Mediated
Communication, 15(2), 314-335. doi:10.1111/j.1083-6101.2010.01522.x
Van Der Heide, B., & Lim, Y. (2015). On the conditional cueing of credibility heuristics:
The case of online influence. Communication Research.
doi:10.1177/0093650214565915
Veil, S., Reynolds, B., Sellnow, T. L., & Seeger, M. W. (2008). CERC as a theoretical
framework for research and practice. Health Promotion Practice, 9(4 suppl), 26S-
34S. doi: 10.1177/1524839908322113
Walther, J. B. (1992). Interpersonal effects in computer-mediated interaction: A relational
perspective. Communication Research, 19, 52–90.
Walther, J. B. (2011). Theories of computer-mediated communication and interpersonal
relations. In M. L. Knapp, & Daly, J. A. (Eds.), Handbook of interpersonal
communication (4th ed., pp. 443-479). Thousand Oaks, CA: Sage.
108
Walther, J. B., & Parks, M. R. (2002). Cues filtered out, cues filtered in: Computer-
mediated communication and relationships. In M. L. Knapp & J. A. Daly (Eds.),
Handbook of interpersonal communication (3rd ed., pp. 529–563). Thousand
Oaks, CA: Sage.
Walther, J. B., Van Der Heide, B., Hamel, L. M., & Shulman, H. C. (2009). Self-
generated versus other-generated statements and impressions in computer-
mediated communication: A test of warranting theory using facebook.
Communication Research, 36(2), 229-253. doi:10.1177/0093650208330251
Walther, J. B., Van Der Heide, B., Tong, S. T., Carr, C. T., & Atkin, C. K. (2010). The
effects of interpersonal goals on inadvertent intrapersonal influence in computer-
mediated communication. Human Communication Research, 36, 323–347.
Waters, R. D., & Williams, J. M. (2011). Squawking, tweeting, cooing, and hooting:
Analyzing the communication patterns of government agencies on twitter.
Journal of Public Affairs, 11, 353–363.
Wendling, C., Radisch, J., & Jacobzone, S. (2013). The use of social media in risk and
crisis communication, OECD working Papers on Public Governance, 24. doi:
10.1787/5k3v01fskp9s-en
Westerman, D., Spence, P.R., & Van Der Heide, B. (2012). A social network as
information: The effect of system generated reports of connectedness on
credibility on twitter. Computers in Human Behavior, 28(1), 199-206. doi:
http://dx.doi.org/10.1016/j.chb.2011.09.001
109
Westerman, D., Spence, P.R., & Van Der Heide, B. (2014). Social media as information
source: Recency of updates and credibility of information. Journal of Computer-
Mediated Communication, 19(2), 171-183. doi: 10.1111/jcc4.12041
Wilson, J. R., & D’Cruz, M. (2006). Virtual and interactive environments for work of the
future. International Journal of Human-Computer Studies, 64(3), 158-169.
Wilson, E. J., & Sherrell, D. L. Source effects in communication and persuasion research:
A meta-analysis of effect size. Journal of the Academy of Marketing Science,
21(2), 101-112. doi:10.1007/bf02894421
110
CURRICULUM VITAE
Xialing Lin
EDUCATION
Western Michigan University M.A. Communication Studies, December, 2012.
Xiamen University B.A. Advertising, June 2007
ACADEMIC POSITIONS
College of Communication & Information, University of Kentucky CIS 300 Course Coordinator Assistant, August, 2015 – May, 2016 Instructor, August, 2014 – Present Graduate Teaching/Research Assistant, 2012 – Present
Ogilvy & Mather, China 2008 – 2009, Account Services in Advertising 2007 – 2008, Public Relations
SCHOLASTIC AND PROFESSIONAL HONORS
Health Communication Fellow, University of Kentucky, Lexington, KY. April 2016.
Post-Graduate Fellow of Communication & Social Robotics Labs, Western Michigan University, Kalamazoo, MI. March, 2016.
Risk/Crisis Communication Fellow, University of Kentucky, Lexington, KY. April 2015.
Top Paper, Communication and Technology Interest Group, the 105th Annual Eastern Communication Association Convention, Providence, RI. April 2014.
Top Paper, Public Relations Division, National Communication Association's 99th Annual Convention, Washington, DC. November 2013.
Top Paper Panel, Communication Theory Division, Central States Communication Association, Kansas City, MO. April 2013.
University First-Class Scholarship, Xiamen University, China. 2006-2007.
Outstanding Print Advertising Award, National Advertising Association, China. 2006.
Outstanding Print Advertising Award, Fujian Advertising Association, China. 2005.
University Distinguished Leaders, Xiamen University, China. 2005-2006.
University Merit Students, Xiamen University, China. 2003-2004.
111
JOURNAL ARTICALES
Spence, P.R., Lachlan, K.A., Westerman, D., Lin, X., Gentile, C. J., Sellnow, T.L., & Sellnow- Richmond, D.D. (accepted). Exemplification effects: Responses to perceptions of risk. Journal of Risk Research.
Lachlan, K. A., Spence, P. R., Lin, X., Najarian, K., & Del Greco, M. (2016). Social media and crisis management: CERC, search strategies, and twitter content. Computers in Human Behavior, 54, 647-652. doi: 10.1016/j.chb.2015.05.027
Spence, P.R., Lachlan, K.A., Lin, X., Sellnow-Richmond, D.D., & Sellnow, T.L. (2015). The problem with remaining silent: Exemplification effects and public image. Communication Studies, 341-357. doi: 10.1080/10510974.2015.1018445
Spence, P. R., Lachlan, K. A., Lin, X., & Del Greco, M. (2015). Variability in Twitter content across the stages of a natural disaster: Implications for crisis communication. Communication Quarterly, 63(2), 171-186. doi: 10.1080/01463373.2015.1012219
Westerman, D., Spence, P.R., & Lin, X. (2014). Telepresence and exemplification in health messages: The relationships among spatial and social presence and exemplars and exemplification effects. Communication Reports, 28(2), 92-102. doi: 10.1080/08934215.2014.971838
Lachlan, K. A., Spence, P. R., & Lin, X. (2014). Expressions of risk awareness and concern through Twitter: On the utility of using the medium as an indication of audience needs. Computers in Human Behavior, 35, 554-559. doi: 10.1016/j.chb.2014.02.029
Lachlan, K.A., Spence, P.R., Lin, X., & Del Greco, M. (2014). Screaming into the wind: Examining the volume and content of tweets associated with hurricane sandy. Communication Studies, 65(5), 500-518. doi: 10.1080/10510974.2014.956941
Lachlan, K.A., Spence, P.R., Lin, X., Najarian, K.M., & Greco, M.D. (2014). Twitter use during a weather event: Comparing content associated with localized and nonlocalized hashtags. Communication Studies, 65(5), 519-534. doi: 10.1080/10510974.2014.956940
Spence, P. R., Lachlan, K. A., Spates, S. A., & Lin, X. (2013). Intercultural differences in responses to health messages on social media from spokespeople with varying levels of ethnic identity. Computers in Human Behavior, 29, 1255-1259.
Spence, P. R., Lachlan, K. A., Spates, S. A., Lin, X., Shelton, A. K., & Gentile, C. J. (2013). Exploring the impact of ethnic identity through other-generated cues on perceptions of spokesperson credibility. Computers in Human Behavior, 29, 3-11.
BOOK CHAPTERS
Lin, X, Xu, Z, Rainer, A., Rice, R., Spence, P. R., & Lachlan, K. A. (accepted). Research in crises: Data collection suggestions and practices. In Eiras, J. R. (Ed), Data collection: Methods, ethical issues and future directions. New York, NY: Nova Science.
112
Lachlan, K.A., Spence, P.R., & Lin, X. (2015). Natural disasters, Twitter, and stakeholder communication: What we know and directions for future research. In L. Austin and J. Yan (Eds.), Social media and crisis communication. Oxford, UK: Routledge.
Spence, P. R., Lachlan, K. A., & Lin, X. (2014). Using the right cues: Directions and implications for communication of health related information through social media. In M. Eaves (Ed.), Applications in health communication: Emerging trends (pp. 205-215). Dubuque, IA: Kendal Hunt.
Lachlan, K.A., Spence, P.R., & Lin, X. (2012). Self-efficacy and learning processes associated with the elderly during disasters and crises. In B. Raskovic and S. Mrdja (Eds.), Natural disasters: prevention, risk factors and management (pp. 327.338). New York, NY: Nova Science.
ENCYCLOPEDIA ENTRIES
Lin, X., & Spence, P. R. (2014). Twitter and public health. In T. Thompson (Ed.), Encyclopedia of health communication. (Vol. 19, pp. 1421-1423). Thousand Oaks, CA: SAGE Publications, Inc. doi: 10.4135/9781483346427.n562
Lin, X., & Spence, P. R. (2013). Spaceflight. In K. Penuel, M. Statler, & R. Hagen (Eds.), Encyclopedia of crisis management. (pp. 891-893). Thousand Oaks, CA: SAGE Publications, Inc. doi: 10.4135/9781452275956.n309
Lin, X., & Spence, P. R. (2013). Reciprocal site. In K. Penuel, M. Statler, & R. Hagen (Eds.), Encyclopedia of crisis management. (pp. 791-793). Thousand Oaks, CA: SAGE Publications, Inc. doi: 10.4135/9781452275956.n271
Lin, X., & Spence, P. R. (2013). Mining. In K. Penuel, M. Statler, & R. Hagen (Eds.), Encyclopedia of crisis management. (pp. 623-625). Thousand Oaks, CA: SAGE Publications, Inc. doi: 10.4135/9781452275956.n211
Spence, P. R., & Lin, X. (2013). Normal accident theory. In K. Penuel, M. Statler, & R. Hagen (Eds.), Encyclopedia of crisis management. (pp. 674-677). Thousand Oaks, CA: SAGE Publications, Inc. doi: 10.4135/9781452275956.n230
113