Wei Zhong
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Wei Zhong has a total of 22 co-authors. Here are the top co-authors by number of shared publications:
- David Andre Broniatowski
The George Washington University
3 shared publications - Jiyoun Suk
University of Connecticut
3 shared publications - Josephine Lukito
Digital Democracy Centre, University of Southern Denmark, Odense, Denmark
3 shared publications - Yunkang Yang
Texas A&M University
3 shared publications - Bin Chen
School of Future Media, The University of Hong Kong, Hong Kong SAR, China
2 shared publications - Jason Greenfield
Department of Computer Science, Princeton University, NJ, USA
2 shared publications
Latest Publications
- 23 September 2026
- Published by SAGE Publications in American Politics Research
- Vol. 55 (1) , 26-42
From Policy to Personality: How Americans Evaluate Presidential Candidates, 1984–2020
This study examines how American voters’ articulated standards for evaluating presidential candidates have changed over time. Using open-ended “likes” and “dislikes” from the American National Election Studies (ANES) from 1984 to 2020, I analyze the criteria citizens invoke and how broadly they evaluate nominees in their own words. I extend ANES’s likes/dislikes coding – which classifies responses into nine evaluative categories spanning policy, party, candidate traits, governance, and other considerations – to recent elections using new human labels and a BERT-based classifier, enabling consistent classification across the full time series. The analysis focuses on three dimensions of expressed evaluation: evaluative content, evaluative breadth, and cross-party evaluative overlap. The results reveal a long-run transformation in articulated electoral judgment. Personal qualities expand sharply to become the dominant evaluative dimension, while policy-based considerations – especially domestic policy – decline in relative prominence. At the same time, evaluative breadth contracts, with respondents articulating fewer distinct criteria per candidate even after accounting for respondent characteristics and interview mode. Finally, cross-party comparison narrows as shared evaluative standards decline and become increasingly concentrated in personal qualities rather than policy. A party-evaluation placebo shows that these patterns are not reducible to generalized open-ended survey fatigue. These patterns indicate that contemporary articulated candidate evaluation is thinner, more personalized, and less anchored in shared programmatic benchmarks, suggesting a narrowing foundation for commensurable electoral accountability in American presidential elections.
- 23 September 2026
- Published by SAGE Publications in American Politics Research
- Vol. 55 (1) , 26-42
From Policy to Personality: How Americans Evaluate Presidential Candidates, 1984–2020
This study examines how American voters’ articulated standards for evaluating presidential candidates have changed over time. Using open-ended “likes” and “dislikes” from the American National Election Studies (ANES) from 1984 to 2020, I analyze the criteria citizens invoke and how broadly they evaluate nominees in their own words. I extend ANES’s likes/dislikes coding – which classifies responses into nine evaluative categories spanning policy, party, candidate traits, governance, and other considerations – to recent elections using new human labels and a BERT-based classifier, enabling consistent classification across the full time series. The analysis focuses on three dimensions of expressed evaluation: evaluative content, evaluative breadth, and cross-party evaluative overlap. The results reveal a long-run transformation in articulated electoral judgment. Personal qualities expand sharply to become the dominant evaluative dimension, while policy-based considerations – especially domestic policy – decline in relative prominence. At the same time, evaluative breadth contracts, with respondents articulating fewer distinct criteria per candidate even after accounting for respondent characteristics and interview mode. Finally, cross-party comparison narrows as shared evaluative standards decline and become increasingly concentrated in personal qualities rather than policy. A party-evaluation placebo shows that these patterns are not reducible to generalized open-ended survey fatigue. These patterns indicate that contemporary articulated candidate evaluation is thinner, more personalized, and less anchored in shared programmatic benchmarks, suggesting a narrowing foundation for commensurable electoral accountability in American presidential elections.
- 5 September 2026
- Published by SAGE Publications in New Media & Society
Cross-platformization: How U.S. right-leaning media curate their posts on Twitter and Truth Social
This study examines how seven right-leaning media organizations utilize cross-platformization strategies to reach audiences across multiple platforms; in this case, Twitter and Truth Social. Focusing on both content volume and story packaging, we collect right-leaning media organizations’ social media posts and content from the 2022 U.S. midterm election. Combining computational and qualitative methods, we find that outlets posted only a small share of total stories across the two platforms and were selective in their content, often posting only on Twitter or Truth Social. For content posted on both platforms, the outlets posted content that ideologically aligned with the users of that platform. More specifically, right-leaning outlets on Truth Social tailored their posts to ideologically aligned users.
- 5 September 2026
- Published by SAGE Publications in New Media & Society
Cross-platformization: How U.S. right-leaning media curate their posts on Twitter and Truth Social
This study examines how seven right-leaning media organizations utilize cross-platformization strategies to reach audiences across multiple platforms; in this case, Twitter and Truth Social. Focusing on both content volume and story packaging, we collect right-leaning media organizations’ social media posts and content from the 2022 U.S. midterm election. Combining computational and qualitative methods, we find that outlets posted only a small share of total stories across the two platforms and were selective in their content, often posting only on Twitter or Truth Social. For content posted on both platforms, the outlets posted content that ideologically aligned with the users of that platform. More specifically, right-leaning outlets on Truth Social tailored their posts to ideologically aligned users.
- 4 March 2026
- Published by Taylor & Francis in Political Communication
- Vol. 43 (2) , 175-191
Political Communication Special Issue Introduction: Multi-Platform Research
Digital media platforms shape and structure how users produce, consume, and engage with content (Nieborg & Poell, 2018) on the internet. In this Special Issue, we define digital media platforms as ...
- 4 March 2026
- Published by Taylor & Francis in Political Communication
- Vol. 43 (2) , 175-191
Political Communication Special Issue Introduction: Multi-Platform Research
Digital media platforms shape and structure how users produce, consume, and engage with content (Nieborg & Poell, 2018) on the internet. In this Special Issue, we define digital media platforms as ...
- Open Access
- 29 October 2025
- Published by Center for Open Science in SocArXiv
It Works When It Works: Measuring the Direct and Indirect Effects of AI Labels on Political Images
Recent advancements in the availability and sophistication of generative artificial intelligence has been accompanied by widespread concerns regarding the ability for the public to navigate the digital information environment, especially during social and political events. A growing consensus across policymakers, academics, and industry leaders has emerged around the need for applying labels communicating whether AI was used in content creation. Labeling as a media literacy strategy and policy intervention has gained momentum, but what impacts do synthetic content labels have on the public? We run two online experiments to measure the effect of content labeling on perceptions of political images—their perceived provenance, veracity, and engagement intention. We find that AI labels effectively communicate content provenance, significantly reducing perceived human involvement in image creation regardless of whether images were actually AI-generated. At the same time, labels have no impact on perceived veracity or self-reported engagement intentions. In a follow on study, we find limited evidence of the "implied authenticity effect," whereby exposure to labeled synthetic images increases the perceived human provenance of subsequent unlabeled synthetic images. However, we show that the near-zero total effect is the result of two offsetting pathways: exposure to labeled synthetic images implies subsequent unlabeled synthetic images are more authentic, but this effect is countered by an increased skepticism caused by prior exposure to labeled content. Taken together, our research contributes to the growing academic literature on content labeling as a media literacy intervention, as well as to ongoing efforts to develop evidence-based strategies to mitigate the harmful political effects of generative AI.
- 29 October 2025
- Published by Center for Open Science in SocArXiv
Open AccessIt Works When It Works: Measuring the Direct and Indirect Effects of AI Labels on Political Images
Recent advancements in the availability and sophistication of generative artificial intelligence has been accompanied by widespread concerns regarding the ability for the public to navigate the digital information environment, especially during social and political events. A growing consensus across policymakers, academics, and industry leaders has emerged around the need for applying labels communicating whether AI was used in content creation. Labeling as a media literacy strategy and policy intervention has gained momentum, but what impacts do synthetic content labels have on the public? We run two online experiments to measure the effect of content labeling on perceptions of political images—their perceived provenance, veracity, and engagement intention. We find that AI labels effectively communicate content provenance, significantly reducing perceived human involvement in image creation regardless of whether images were actually AI-generated. At the same time, labels have no impact on perceived veracity or self-reported engagement intentions. In a follow on study, we find limited evidence of the "implied authenticity effect," whereby exposure to labeled synthetic images increases the perceived human provenance of subsequent unlabeled synthetic images. However, we show that the near-zero total effect is the result of two offsetting pathways: exposure to labeled synthetic images implies subsequent unlabeled synthetic images are more authentic, but this effect is countered by an increased skepticism caused by prior exposure to labeled content. Taken together, our research contributes to the growing academic literature on content labeling as a media literacy intervention, as well as to ongoing efforts to develop evidence-based strategies to mitigate the harmful political effects of generative AI.
- Open Access
- 21 August 2025
- Published by Cambridge University Press (CUP) in Perspectives on Politics
- p. 1-26
Subdued but Unbroken: The Cohesion of Far-Right Extremist Followers after Deplatforming
While deplatforming has become an increasingly common strategy to combat online harm and far-right extremism, its effects on the followers of extremist groups—who are key supporters and play a crucial role in spreading and sustaining these ideologies—remain underexplored. On August 10, 2018, Twitter (now X) deplatformed one such far-right extremist group, the Proud Boys, along with their affiliated accounts. Leveraging this intervention, our research addresses a key knowledge gap by examining the impact of deplatforming on the cohesion of extremist group followers. Specifically, we investigate whether deplatforming leads to fragmentation or reinforces unity among the group’s followers. We assess cohesion through three theoretical lenses: task commitment, social commitment, and sense of belonging. By analyzing over 12 million tweets from approximately nine thousand Proud Boys supporters between August 1, 2017, and September 1, 2019, we find that deplatforming had a limited effect on reducing group cohesion. Instead, it may have prompted followers to seek broader networks and external interactions, leaving overall cohesion largely intact. This study offers important insights into the resilience of online extremist communities and the limitations of deplatforming as a strategy to disrupt them. Understanding these dynamics is essential for developing more effective approaches to counter online extremism and promote safer digital spaces.
- 21 August 2025
- Published by Cambridge University Press (CUP) in Perspectives on Politics
- p. 1-26
Open AccessSubdued but Unbroken: The Cohesion of Far-Right Extremist Followers after Deplatforming
While deplatforming has become an increasingly common strategy to combat online harm and far-right extremism, its effects on the followers of extremist groups—who are key supporters and play a crucial role in spreading and sustaining these ideologies—remain underexplored. On August 10, 2018, Twitter (now X) deplatformed one such far-right extremist group, the Proud Boys, along with their affiliated accounts. Leveraging this intervention, our research addresses a key knowledge gap by examining the impact of deplatforming on the cohesion of extremist group followers. Specifically, we investigate whether deplatforming leads to fragmentation or reinforces unity among the group’s followers. We assess cohesion through three theoretical lenses: task commitment, social commitment, and sense of belonging. By analyzing over 12 million tweets from approximately nine thousand Proud Boys supporters between August 1, 2017, and September 1, 2019, we find that deplatforming had a limited effect on reducing group cohesion. Instead, it may have prompted followers to seek broader networks and external interactions, leaving overall cohesion largely intact. This study offers important insights into the resilience of online extremist communities and the limitations of deplatforming as a strategy to disrupt them. Understanding these dynamics is essential for developing more effective approaches to counter online extremism and promote safer digital spaces.
- Open Access
- 1 April 2025
- Published by SAGE Publications in Social Media + Society
- Vol. 11 (2)
Candidates Be Posting: Multi-Platform Strategies and Partisan Preferences in the 2022 U.S. Midterm Elections
In this multi-platform, comparative study, we analyze social media messages from political candidates ( N = 1,517) running for Congress during the 2022 U.S. Midterm election. We collect data from seven social media platforms: Facebook, Twitter, Truth Social, Gettr, Instagram, YouTube, and Rumble over the 4 weeks before and after election day. With this unique dataset of posts, we apply computational methods to identify messages that sought to mobilize individuals (online and offline) to donate money, vote, attend events, engage with the campaign online, and visit the campaign’s content on other platforms. We find that Democrats were not on alt-tech platforms in 2022 and that both Republicans and Democrats use video-based platforms for multiple mobilization strategies. Mobilization messages varied for House and Senate candidates of both parties across platforms, before and after election day.
- 1 April 2025
- Published by SAGE Publications in Social Media + Society
- Vol. 11 (2)
Open AccessCandidates Be Posting: Multi-Platform Strategies and Partisan Preferences in the 2022 U.S. Midterm Elections
In this multi-platform, comparative study, we analyze social media messages from political candidates ( N = 1,517) running for Congress during the 2022 U.S. Midterm election. We collect data from seven social media platforms: Facebook, Twitter, Truth Social, Gettr, Instagram, YouTube, and Rumble over the 4 weeks before and after election day. With this unique dataset of posts, we apply computational methods to identify messages that sought to mobilize individuals (online and offline) to donate money, vote, attend events, engage with the campaign online, and visit the campaign’s content on other platforms. We find that Democrats were not on alt-tech platforms in 2022 and that both Republicans and Democrats use video-based platforms for multiple mobilization strategies. Mobilization messages varied for House and Senate candidates of both parties across platforms, before and after election day.
- Open Access
- 13 February 2024
- Published by Cambridge University Press (CUP) in American Political Science Review
- Vol. 118 (4) , 2054-2071
“Keep Your Heads Held High Boys!”: Examining the Relationship between the Proud Boys’ Online Discourse and Offline Activities
This study examines the relationship between online communication by the Proud Boys and their offline activities. We use a supervised machine learning model to analyze a novel dataset of Proud Boys Telegram messages, merged with US Crisis Monitor data of violent and nonviolent events in which group members participated over a 31-month period. Our analysis finds that intensifying expressions of grievances online predict participation in offline violence, whereas motivational appeals to group pride, morale, or solidarity share a reciprocal relationship with participation in offline events. This suggests a potential online messaging–offline action cycle, in which (a) nonviolent offline protests predict an increasing proportion of motivational messaging and (b) increases in the frequency and proportion of motivational appeals online, in turn, predict subsequent violent offline activities. Our findings offer useful theoretical insights for understanding the relationship between online speech and offline behavior.
- 13 February 2024
- Published by Cambridge University Press (CUP) in American Political Science Review
- Vol. 118 (4) , 2054-2071
Open Access“Keep Your Heads Held High Boys!”: Examining the Relationship between the Proud Boys’ Online Discourse and Offline Activities
This study examines the relationship between online communication by the Proud Boys and their offline activities. We use a supervised machine learning model to analyze a novel dataset of Proud Boys Telegram messages, merged with US Crisis Monitor data of violent and nonviolent events in which group members participated over a 31-month period. Our analysis finds that intensifying expressions of grievances online predict participation in offline violence, whereas motivational appeals to group pride, morale, or solidarity share a reciprocal relationship with participation in offline events. This suggests a potential online messaging–offline action cycle, in which (a) nonviolent offline protests predict an increasing proportion of motivational messaging and (b) increases in the frequency and proportion of motivational appeals online, in turn, predict subsequent violent offline activities. Our findings offer useful theoretical insights for understanding the relationship between online speech and offline behavior.
- Open Access
- 13 July 2023
- Published by Center for Open Science in SocArXiv
Evaluating Twitter's COVID-19 Vaccine Misinformation Removal Policy
Objectives. To assess the efficacy of Twitter's March 1, 2021 COVID-19 vaccine misinformation removal policy.Methods. We collected over 400 million English-language tweets related to COVID-19 using more than 100 pertinent keywords from February 6, 2020, to December 15, 2022, comparing vaccine-related tweets and corresponding accounts before vs. after the implementation of Twitter's interventions. We used a comparative interrupted time series analytic approach to compared the content of misinformative (>10% of tweets contain links to misinformative websites) to non-misinformative accounts. Results. We identified 7,084 misinformative accounts and 6,706,999 non-misinformative accounts. Misinformative accounts were 1.43 times more likely to persist (RR=1.43; 95% CI:1.41-1.45; P<0.001) and 30.16 times more likely to become more misinformative (RR=30.16; 95% CI: 28.48 - 31.95; P<0.001) compared to non-misinformative accounts. We did not detect a significant decrease in content from misinformative accounts (RR=0.03; 95% CI: 0.00 - 3.93; P=0.16) compared to pre-policy data; however, we did detect a significant decrease in content from non-misinformative accounts (RR=0.02; 95% CI: 0.00 - 0.32; P=0.005). We did not detect a significant difference between these two groups (RR=1.48; 95% CI: 0.01 - 398.61; P=0.89). Conclusion. Twitter's vaccine misinformation removal policies do not appear to have been associated with a detectable reduction in content from misinformative, compared to non- misinformative, users or in more misinformative compared to less misinformative tweets.Public Health Implications. Our study highlights the necessity for external evaluations to ascertain the sufficiency of platforms’ self-regulatory measures to safeguard public health.
- 13 July 2023
- Published by Center for Open Science in SocArXiv
Open AccessEvaluating Twitter's COVID-19 Vaccine Misinformation Removal Policy
Objectives. To assess the efficacy of Twitter's March 1, 2021 COVID-19 vaccine misinformation removal policy.Methods. We collected over 400 million English-language tweets related to COVID-19 using more than 100 pertinent keywords from February 6, 2020, to December 15, 2022, comparing vaccine-related tweets and corresponding accounts before vs. after the implementation of Twitter's interventions. We used a comparative interrupted time series analytic approach to compared the content of misinformative (>10% of tweets contain links to misinformative websites) to non-misinformative accounts. Results. We identified 7,084 misinformative accounts and 6,706,999 non-misinformative accounts. Misinformative accounts were 1.43 times more likely to persist (RR=1.43; 95% CI:1.41-1.45; P<0.001) and 30.16 times more likely to become more misinformative (RR=30.16; 95% CI: 28.48 - 31.95; P<0.001) compared to non-misinformative accounts. We did not detect a significant decrease in content from misinformative accounts (RR=0.03; 95% CI: 0.00 - 3.93; P=0.16) compared to pre-policy data; however, we did detect a significant decrease in content from non-misinformative accounts (RR=0.02; 95% CI: 0.00 - 0.32; P=0.005). We did not detect a significant difference between these two groups (RR=1.48; 95% CI: 0.01 - 398.61; P=0.89). Conclusion. Twitter's vaccine misinformation removal policies do not appear to have been associated with a detectable reduction in content from misinformative, compared to non- misinformative, users or in more misinformative compared to less misinformative tweets.Public Health Implications. Our study highlights the necessity for external evaluations to ascertain the sufficiency of platforms’ self-regulatory measures to safeguard public health.
- 7 December 2022
- Published by Elsevier in Social Science & Medicine
- Vol. 317, 115594
Economic risk framing increases intention to vaccinate among Republican COVID-19 vaccine refusers
- 7 December 2022
- Published by Elsevier in Social Science & Medicine
- Vol. 317, 115594
Economic risk framing increases intention to vaccinate among Republican COVID-19 vaccine refusers