The Effects of Engagement of Adolescents’ Mothers with Social Media Anti- and Pro-Vaccination Content on Their Children’s Human Papillomavirus Vaccine Uptake with Vaccine Hesitancy as the Mediator
Objectives: (1) To compare the effects of engagement of adolescents’ mothers with anti- and pro-vaccination messages on their children’s HPV-vaccination rates, and (2) to identify the role of vaccine hesitancy that mediates such relationships.
Methods: In December 2019, we conducted a survey among 426 US mothers of adolescents aged 13 to 18. For data analyses, we employed logistic regression for the HPV-vaccine initiation and completion, and ordinary least squares regression for adhering to the recommended schedule.
Results: Mothers’ engagement with antivaccination messages is significantly associated with decreases in their children’s HPV-vaccination rates, while their engagement with pro-vaccination content is not linked to increases in the rates. Vaccine hesitancy mediates the impact of engagement with antivaccination content on the rates.
Conclusions: Using the notion of loss-aversion in Behavioral Economics, we demonstrate how mothers of adolescents become more susceptible to antivaccination messages than to pro- content and how vaccine hesitancy towards the broad childhood vaccine system leads to the stagnant uptake of the vaccine.
Policy Implications: Policy makers should consider suppressing the propagation of antivaccination messages vis social media among mothers, because spreading pro-vaccination messages alone will not be effective and because mothers are the key HPV-vaccination decision-makers.
Acknowledgement: This study was funded by the Launch Award from the Diversity Research Network at Michigan State University.
Young Anna Argyris
Human-in-the-Loop Development of Rapid Intelligent Misinformation Detector for Social Media
Identification of user resistance tactics to algorithmic regulation: Our conceptual framework, Resistance to Algorithm Model, delineates how anti-vaccinists attempt to circumvent algorithmic using multimodal and pseudo-scientific tactics, from our qualitative analyses of longitudinal, cross-site observations. The research team builds a large dataset (~1.8 million) pertinent to vaccine as infrastructure to facilitate future efforts.
Development of technically novel approaches to translating humans’ input seamlessly and cost-effectively into Artificial Intelligence development and refinement: Technological innovations, such as multimodal detection, retraining Artificial Intelligence with a decreasing amount of human input, and real-time adaptation to emerging new tactics, will be implemented.
Evaluation of Artificial Intelligence efficacy in realistic interactions with users: A novel evaluation mechanism of Artificial Intelligence will be produced that tests its abilities to detect new, emerging instances non-existent in the training set, and its impact on users.
Our detector will adapt to adversaries’ morphing resistance tactics, thereby suppressing the propagation of anti-vaccine messages and preventing unsuspecting individuals from making detrimental decisions. In addition, the same method that we propose for our human-in-the-loop AI development can be modified for other topics of misinformation content in multiple modalities (e.g., texts, static pictures and animated videos) on multiple platforms, including various social media sites (Facebook, Instagram, Youtube, and Twitter), web search engines, and individual/organizational desktop and mobile websites. Thus, our outcomes will have broad and positive societal impacts.
JBHI: Dr. Instagram May Be a Liar: Detecting Medical Misinformation on Social Media
The Effects of Visual Congruence on Increasing Consumers' Brand Engagement: An Empirical Investigation of Influencer Marketing on Instagram Using Deep-Learning Algorithms for Automatic Image Classification
Get Online Access:
Citation: Argyris, Young, Wang, Zuhui, Kim, Yongsuk & Yin, Zhaozheng. (2020). The effects of visual congruence on increasing consumers’ brand engagement: An empirical investigation of influencer marketing on Instagram using deep-learning algorithms for automatic image classification. Computers in Human Behavior, 112, 106443. doi:10.1016/j.chb.2020.106443.
Background: Influencers are non-celebrity individuals who gain popularity on social media by posting visually attractive content (e.g., photos and videos) and by interacting with other users (i.e., Followers) to create a sense of authenticity and friendship. Brands partner with Influencers to garner engagement from their target consumers in a new marketing strategy known as "Influencer marketing." Nonetheless, the theoretical underpinnings of such remains unknown.
Objectives: We suggest a new conceptual framework of "Visual-Congruence-induced Social Influence (VCSI)," which contextualizes the Similarity-Attraction Model in the Social Influence literature. Using VCSI, we delineate how Influencers use visual congruence as representations of shared interests in a specific area to build strong bonds with Followers. This intimate affiliation catalyzes (i.e., mediates) the positive effects of visual congruence on Followers' brand engagement.
Methods: To test these hypotheses, we conducted in vivo observations of Influencer marketing on Instagram. We collected >45,000 images and social media usage behaviors over 26 months. We then applied deep-learning algorithms to automatically classify each image and used social media analytics to disclose hidden associations between visual elements and brand engagement.
Results: Our hypothesis testing results provide empirical support for VCSI. Specifically, our results show the following:
Social media Influencers post visual content congruent with Followers’ interests.
Visual congruence increases followers’ engagement with Influencers’ posts.
Such an increase in turn augments followers’ engagement with the endorsed brand.
Affiliation between Influencers and Followers mediates the above relationships.
Theoretical Contributions: We propose a new conceptual framework, VCSI, that contextualizes Similarity Attraction Model to Social Influence. This new framework expands Similarity Attraction Model by incorporating multimodal elements of social media posts and by delineating the processes by which Influencers affect their Followers’ brand engagement. We have also provided systematic and in vivo observations regarding the effectiveness of Influencers to increase brand engagement among large crowds. This rare real-world data collection and observations warrant adequate empirical grounds for advancing theories.
Practical Implications: Images can create substantial social influence for message recipients on social media. Influencers and brand managers alike should recognize the power of images in their messages and should approach them more strategically. Simultaneously, with increased intimacy and frequent interactions, Influencers can succeed at garnering Followers’ engagement with the brand. Lastly, visual congruence must be built within a brand-pertinent niche area for Followers to engage with the brand.
Developing a Vaccine Informatics to Identify Message Frames Used in Vaccine Debates on Social Media: Combining Automatic Tweet Classification and Clustering Machine-Learning Algorithms with Qualitative Content Analysis
Background: Exposure to anti-vaccine content on social media has been associated with delays and refusals of vaccinations, while pro-vaccine campaigns devised to disseminate the preventive benefits of vaccines have not succeeded in increasing vaccine uptake rates. Reasons remain unknown why anti-vaccine messaging hampers uptake while pro-vaccine campaigns do not improve it.
Objective: We aim to identify reasons for the disparate effectiveness of anti- versus pro-vaccine social media content on vaccine delivery rates. In so doing, we apply the perspectives of message framing used in interpersonal health communication to explain why individuals refuse to adopt preventive behaviors. Specifically, we compare (1) the diversity, coherence, and distinctiveness of topics discussed by pro- and anti-vaccine communities and (2) message frames used to portray vaccines as a public health accomplishment or harmful agents.
Methods: We developed a multimethod that combines the collection of a large amount of data from Twitter (~40,000 tweets), an automatic tweet classification algorithm, the K-means clustering algorithm, and a qualitative content analysis.
Results: Our results show a larger number of topics (20 versus 17 clusters), greater coherence of topics (0.99 vs. 0.97) and distinctiveness of topics (1.22 vs. 1.31) among anti-vaccinists in comparison to pro-vaccinists. In addition, while anti-vaccinists use all four message frames known to make narratives persuasive and influential, pro-vaccinists neglect the problem statement.
Conclusions: Based on our results, we attribute the diversity, coherence, and distinctiveness of topics discussed among anti-vaccinists to their higher engagement, and we ascribe the influence of vaccine debate on uptake rates to the comprehensiveness of the message frames. These results show the urgency of developing value propositions for vaccines to counteract the negative impact of anti-vaccine content on the uptake rates.
Trial Registration: This study was determined to be a non-human subject study by Michigan State University’s Institutional Review Board (#STUDY00004514).
JMIR: Developing a Vaccine Informatics to Identify Message Frames Used in Vaccine Debates on Social Media: Combining Automatic Tweet Classification and Clustering Machine-Learning Algorithms with Qualitative Content Analysis
Young Anna Argyris
The Effects of the Visual Presentation of an Influencer’s Extroversion on Perceived Credibility and Purchase Intentions—Moderated by Personality Matching with the Audience
Background: Influencers are ordinary individuals who have amassed large followings by demonstrating expertise in various niches on social media sites. We chose the source credibility model and the similarity-attraction model as our theoretical frameworks to identify what created their social influence.
Objectives: We aim to propose (1) the visual presentation of an Influencer’s extroversion as an antecedent to source credibility and purchase intentions and (2) personality matching in terms of extroversion between an Influencer and their audience as a moderator of such relationships.
Methodology: In our controlled online experiment (n = 165), the profile of a brand ambassador of a fashion brand was curated to create two levels of extroversion (i.e. low vs. high extroversion). Participants’ self-reported extroversion levels were counterbalanced in these two conditions to create a personality match with the Influencer, entitled “extroversion congruence.”
Results: Our results suggest that the visual presentation of an Influencer’s extroversion increases the perceived credibility of the Influencer and subsequent purchase intentions. Additionally, our findings show that these relationships were asymmetrically moderated by extroversion congruence: the positive effects of extroversion increase in the case of high extroversion among both the Influencer and their audience but decrease in the case of low extroversion.
Conclusion: Personality match between influencers and their audiences, especially in the case of high extraversion portrayed in visual social media posts, augment the influencers’ impact on their audiences decision-making and behaviors.
Young Anna Argyris
Using Speech Acts for Complaint Management on Social Media
Background: A carefully tailored tone in response to a complaint on social media can relieve customers’ negative emotions. However, very few studies have identified what response tones, based on an established theory, would be most effective for appeasing upset customers.
Objectives: This study conceptualizes a service agent’s response tones based on Ballmer and Brennenstuhl’s (1981) classification of speech acts and examines how an agent’s use of speech acts alleviate complainants’ negative emotions. Ballmer and Brennenstuhl classify speech acts within the dimensions of conventionality and dialogicality, and they suggest the two dimensions interact. Thus, we examine the impact of each dimension of speech acts and the interactions between the two dimensions on the alleviation of complainants’ negative emotions.
Methodology: We collected >160,000 tweets showing 34 companies’ complaint handling cases. Company agents’ response tones were classified using our deep learning algorithms (i.e., bi-directional recurrent neural networks).
Results: Our fixed-effect regression results show that a low level of each speech acts alleviates complainants’ emotions but that the combination of the two erodes the individual advantages.
Conclusion: This study expands Ballmer and Brennenstuhl’s (1981) speech act classification from a speaker’s perspectives to a listener’s perspectives, by contextualizing it in an analysis of service agents’ tones and their roles in relieving complainants’ negative emotions.
Young Anna Argyris
Literature Review of Social Media Interventions about Vaccines
Methods: This review searched for relevant systematic review articles that have been published in peer-reviewed, scholarly journals since 2010. Seven databases were searched: PubMed, PsycINFO, Communication & Mass Media Complete, Sociological Abstracts, Web of Science, EBSCO Academic Search Complete, and Educational Resource Information Center (ERIC). Keywords were derived from the themes of social media, campaigns, and vaccination using the National Institute of Health’s National Library of Medicine Medical Subject Headings (MeSH). Searched terms were used in combination with Boolean operators to yield studies. The database searches generated 104 articles.
Young Anna Argyris