YXYiwei Xu

Yiwei Xu

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Yiwei Xu has a total of 23 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    AI Overviews Reduce Search Result Engagement & May Influence Attitude Polarization

    by Saloni Dash, Yiwei Xu, Lei Cai, Sterling Williams-Ceci, Wang Liao, Emma Spiro

    AI-generated overviews (AIOs) increasingly appear at the top of search result pages, making them the first, and potentially only, content users encounter during search. Despite their widespread adoption, their impact on information seeking, processing, and downstream attitudes remains critically underexplored. Across two preregistered experiments (N = 2,797), we randomly assigned participants to view AIOs on one of four contested topics manipulated to be attitude-congruent, attitude-incongruent, or balanced (only Study 2) relative to their prior attitudes. Participants who viewed an AIO, irrespective of content, were significantly less likely to engage with subsequent search results than control (no AIO). Participants perceived attitude-incongruent AIOs as more biased and weaker than attitude-congruent AIOs. Despite this, attitude-incongruent AIOs reduced attitude polarization relative to attitude-congruent AIOs in Study 1, but not in Study 2 (although directionally consistent). Taken together, these findings suggest that AIOs significantly influence information seeking and processing and raise critical concerns for democratic deliberation.

  • Open Access

    The persuasive potential of AI-paraphrased information at scale

    by Saloni Dash, Yiwei Xu, Madeline Jalbert, Emma S Spiro

    In this article, we study how AI-paraphrased messages have the potential to amplify the persuasive impact and scale of information campaigns. Building from social and cognitive theories on repetition and information processing, we model how CopyPasta—a common repetition tactic leveraged by information campaigns—can be enhanced using large language models. We first extract CopyPasta from two prominent disinformation campaigns in the United States and use ChatGPT to paraphrase the original message to generate AIPasta. We then validate that AIPasta is lexically diverse in comparison to CopyPasta while retaining the semantics of the original message using natural language processing metrics. In a preregistered experiment comparing the persuasive potential of CopyPasta and AIPasta (N = 1,200), we find that AIPasta (but not CopyPasta) is effective at increasing perceptions of consensus in the broad false narrative of the campaign while maintaining similar levels of sharing intent with respect to Control (CopyPasta reduces such intent). Additionally, AIPasta (vs. Control) increases belief in the exact false claim of the campaign, depending on political orientation. However, across most outcomes, we find little evidence of significant persuasive differences between AIPasta and CopyPasta. Nonetheless, current state-of-the-art AI-text detectors fail to detect AIPasta, opening the door for these operations to scale successfully. As AI-enabled information operations become more prominent, we anticipate a shift from traditional CopyPasta to AIPasta, which presents significant challenges for detection and mitigation.


  • Promoting Teen Pregnancy Prevention: An Analysis of Social Media Content Strategy Over Five Years

    by Erin Ash, Emma Cox, Yiwei Xu, Brandon Boatwright

    Although rates in the United States have declined significantly since the 1990s, the U.S. continues to have among the highest teen birth rates in the developed world, and recent policy changes in the U.S. warrant increased attention to pregnancy prevention efforts. This research employed a computational approach to analyze social media messages about a pregnancy prevention campaign with focus on analyzing responsibility attribution and frames that might lead to potential stigmatization. Original posts from the Power to Decide campaign's Twitter account posted from 2017 to 2022 (N = 7,770) were analyzed. Automated topic modeling identified five topics extracted from the data: access, encouraging conversations, pregnancy prevention programs, voter voice, and barriers. A subsequent thematic analysis was conducted to uncover sub-themes that provide insight into each topic. Results revealed an emphasis on social responsibility in the organization's social media messaging, reflecting the multiple stakeholders the organization targets in its social media messaging, which are not limited to the at-risk population (i.e. teens), and to whom structural solutions to teen pregnancy prevention can be promoted.


  • Open Access

    AI-Paraphrasing Increases Perceptions of Social Consensus & Belief in False Information

    by Saloni Dash, Yiwei Xu, Emma Spiro

    Large Language Models (LLMs) have the potential to enhance message features and exploit cognitive heuristics to increase the persuasiveness of strategic information campaigns. In this paper, we focus on one such heuristic cue --- repetition, which is known to increase belief through the illusory truth effect. We investigate repetition by extracting verbatim repetitive messaging (CopyPasta) from recent U.S. disinformation campaigns. After using an LLM to rewrite the CopyPasta messages, we show that AI-paraphrased messages (AIPasta) are lexically diverse in comparison to their CopyPasta counterparts while retaining the semantics of the original message. In a preregistered experiment comparing the persuasive effects of CopyPasta and AIPasta (N = 1,200, U.S. nationally representative sample), we find that AIPasta (versus control) increases perceptions of social consensus for false claims, particularly among participants less familiar with the false narrative. Among Republican participants, AIPasta also increases belief in false claims as compared to the control group. Additionally, AIPasta (versus control) increases sharing intention of the disinformation messages among participants, as well as the degree to which the broader false claims are recalled. Notably, we find that CopyPasta does not demonstrate the same level of persuasiveness across these dimensions. Broadly, our findings suggest that generative AI tools have the potential to amplify or increase the effectiveness of strategic information campaigns through strategies like non-verbatim repetition. Results have implications for the detection and mitigation of harmful and false information.

  • Collective Information Seeking During a Health Crisis : Predictors of Google Trends During COVID-19

    by Yiwei Xu, Drew Margolin

    This article approaches collective health information seeking from computational method by investigating patterns of Google Trends data in the United States during the early stages of the COVID-19 pandemic. We analyzed factors that prompted a community’s curiosity, and information that communities were most curious about. The results of our cross-sectional and time-series-based analyses reveal a few salient findings: (1) Republican leaning states searched less frequently, and while states with more cases searched more, partisan lean is a more significant predictor; (2) States with greater level of poverty searched less frequently; (3) Leadership on the national level significantly influenced people’s searching behavior; (4) Communities were most interested in “local risk” information as well as quantifiable information. We show in this work that established individual information seeking theoretical predictors (risk) can predict online collective information demand and information seeking subcategories with important contributions from collective conditions (leadership). Health communication practitioners can design health messages and choose media channels more purposefully according to what people are most interested in searching.


  • Open Access

    Local TV News Coverage of Racial Disparities in COVID-19 During the First Wave of the Pandemic, March–June 2020

    by Yiwei Xu, Elizabeth K. Farkouh, Caroline A. Dunetz, Sravya L. Varanasi, Sophia Mathews, Sarah E. Gollust, Erika Franklin Fowler, Steven Moore, Neil A. Lewis, Jeff Niederdeppe

    The COVID-19 pandemic has disproportionately impacted health and social outcomes for people of color in the United States. This study examined how local TV news stories attributed causes and solutions for COVID-19-related racial health and social disparities, and whether coverage of such disparities changed after George Floyd’s murder, during the first wave of the COVID-19 pandemic. We systematically validated keywords to extract relevant news content and conducted a content analysis of 169 discrete local TV news stories aired between March and June 2020 from 80 broadcast networks within 22 purposefully selected media markets. We found that social determinants of COVID-19 related racial disparities have been part of the discussion in local TV news, but racism as a public health crisis was rarely mentioned. Coverage of racial disparities focused far more attention on physical health outcomes than broader social impacts. Stories cited more structural factors than individual factors, as causes of these disparities. After the murder of George Floyd, stories were more likely to mention Black and Latinx people than other populations impacted by COVID-19. Only 9% of local news stories referenced racism, and stories referenced politicians more frequently than public health experts.


  • Testing Strategies to Increase Source Credibility through Strategic Message Design in the Context of Vaccination and Vaccine Hesitancy

    by Yiwei Xu, Drew Margolin, Jeff Niederdeppe

    Health communicators in the United States face substantial challenges in their efforts to increase parent uptake of the measles, mumps, and rubella (MMR) vaccine for their children. One major set of challenges involves low levels of trust in medical science behind vaccination safety and effectiveness, pharmaceutical companies who produce these vaccines, and government health agencies who promote vaccination. We conducted a two-wave randomized experiment (N = 1,000 at time 1, t1, N = 803 at time 2, t2) to test whether messages designed to convey the expertise, trustworthiness, or caring/goodwill of a governmental source of information (the Centers for Disease Control and Prevention) increased perceived source credibility, increased parent intentions to vaccinate their children, and/or reduced vaccine hesitancy. We found no support for any of the study’s original, pre-registered hypotheses. However, post-hoc analyses reveal a variety of promising directions for future work on strategic messaging to increase source credibility in the context of vaccine hesitancy. A message designed to convey source expertise produced greater perceived caring/goodwill among parents overall. Furthermore, among parents who were vaccine hesitant at baseline, a message originally designed to convey source expertise produced greater perceived trustworthiness and reduced vaccine hesitancy among this group.


  • Framing Use of Force: An Analysis of News Organizations’ Social Media Posts About Police Shootings

    by Erin Ash, Yiwei Xu, Alexandria Jenkins, Chenjerai Kumanyika

    This research is a systematic investigation of reporting on police use of force incidents by news organizations through social media. A sample of Tweets and Facebook posts appearing on the social media accounts of top news outlets (N = 500) over the course of a year were analyzed. Rationalizations for use of force, characterizations of police and victims, and contextual framing were examined. Results revealed medium-based differences in challenging police actions as well as demonstrated the limitations of short-form social media in communicating news on complex issues. Implications of these findings for theory and journalistic practice are discussed.