Fact-checking agencies are essential to correct misinformation and inform the public, while how people evaluate these agencies and their messages remain unclear (Brandtzaeg et al., 2017). Two factors about the messages and the sources – two essential factors in the theories of persuasion – were examined: language intensity of fact-checking labels and AI as a fact-checking agency. Language intensity, a linguistic feature that reflects message specificity and emotionality, may implicitly influence the acceptance of misinformation corrections and behavior intentions (Bowers, 1963). While AI has the potential to automate the fact-checking process and improve the acceptance of misinformation corrections as an unbiased automated decision maker, the social acceptance of AI in fact checking is unclear. This study investigated how language intensity and fact-checking agency (human vs. AI) influence the evaluations of fact-checking messages and agencies with an observational study of fact-checking messages on social media (N = 33755) and two online experiments (combined N = 1449) in the U.S. Both the observational study and the experiments showed that fact-checking messages with high language intensity would elicit low message credibility, while this effect diminished when the messages were counter-attitudinal in the experiments. Besides, participants perceived AI fact-checking agencies the same as human agencies. Individual differences in conspiracy ideation, political ideology and demographics significantly affected message credibility and engagement intentions as well. These findings suggest that language nuances such as language intensity in fact-checking messages affected message perception and the acceptance of misinformation corrections. Theoretical and practical implications were discussed in detail.

Short video platforms have emerged as key information sources, while the proliferation of misinformation on these platforms poses challenges for science and risk communication to adapt to emerging media formats and provide effective scientific surveillance on these platforms. This dissertation proposes that sensational short videos should be considered as an innovative solution to effectively communicate science and risk information in the current information ecosystem. Grounded on the theoretical framework of Message Sensation Value (MSV), this dissertation (a) investigates the relationship between MSV and public engagement and develops a computational measure of MSV with a computational observational study, and (b) examines the theoretical mechanism by which MSV in short videos impacts message attention, processing, and persuasiveness in the context of COVID-19 and climate change with an online experiment.Relying on a computational analysis of N = 1,200 short videos, the observational study reveals a consistent inverted U-shaped relationship between MSV and engagement, where videos with moderate levels of MSV are most engaging. This study also presents a computational measure of MSV using multimodal features with high reliability and validity. Built upon the insights in the observational study, the online experiment uses a within-subject experimental design and investigates the impacts of MSV (low, mid, high) in short videos for N = 1,467 participants in the United States. Results showed that MSV enhanced message processing and persuasiveness across varying belief congruence, while MSV did not affect attention or risk perception. In addition, contrary to existing research that emphasizes the impact of sensation-seeking and belief congruence, age was found to moderate MSV’s effects, which were particularly prominent for young adults. Overall, with both observational and experimental evidence, this dissertation demonstrates that MSV enhances message engagement and persuasiveness to a certain extent and suggests the possibility of leveraging sensational multimodal features in short videos to improve message persuasiveness and engagement for effective science and risk communication. This dissertation provides significant theoretical contributions and practical implications for communicating risks in the contemporary media landscape. First, this dissertation advances our understanding of MSV’s persuasive impacts and provides a useful theoretical framework for understanding multimodal information. Second, this computational measure of MSV provides a robust and automated tool for estimating MSV in short videos, facilitating the empirical research of short videos and MSV. Lastly, this study provides theory-informed practical implications and actionable guidelines for content production on short video platforms, which could increase the visibility of scientific information and establish scientific surveillance on short video platforms.

Background

Methods

Results

Conclusion

Background

Objective

Methods

Results

Conclusions