Emily K. Vraga

Existing research offers mixed evidence on how social media use contributes to or hinders healthy democratic societies. This study examines the impact of social media use generally, and perceived exposure to political misinformation and correction on social media specifically, on two forms of democratic citizenship: constructive behaviors and destructive democratic attitudes. Drawing on a two-wave survey of US adults, we find that greater social media use is linked to heightened concern about election fairness but not to greater political participation among political partisans. Higher perceived exposure to political misinformation predicts attendance at political events but is unrelated to destructive democratic attitudes. Higher perceived exposure to corrections is positively associated with donating to campaigns and volunteering, but also predicts stronger endorsement of violence for political goals. These results reveal the mixed role of social media in encouraging democratic participation while amplifying some anti-democratic attitudes.

Research demonstrates that correction consistently reduces misperceptions, increasing the importance of understanding how often people see such corrections. To answer this, researchers often rely on self-reported measures through surveys. However, there is little consistency in these measures across studies, hindering comparability and theory development. To investigate whether different measures of self-reported correction experiences affect their estimated frequency, this study uses a preregistered online survey experiment to examine two key factors: the question wording ("corrected" vs. "told they shared misinformation") and the response options (binary vs. four vs. five-point frequency scales). Findings reveal that frequency scales produce consistently higher estimates of three correction experiences (witnessing, performing, and being corrected) compared to binary measures, although classifying those who rarely experience correction alongside those who never do reduces these differences. Moreover, using the term "correction" leads to higher self-reported instances of performing corrections than the "told misinformation" phrasing, but does not impact estimates of witnessing or being corrected. These results offer researchers clear takeaways for comparing measures across multiple datasets and reinforce the importance of moving toward consistent measures in the future to ensure we can better understand how often people experience corrections.

Declining trust in public health organizations has (not coincidentally) coincided with a rise in health misinformation on social media. During the COVID-19 pandemic, the World Health Organization (WHO) strategically employed corrective communications to counter common misinformation as an important aspect of its risk communication and community engagement (RCCE) efforts. Whereas past research focuses mainly on the ability of corrections to reduce misperceptions, we turn attention to the broader question of whether and how they affect perceptions of the WHO. We analyze an online survey experiment (N = 1343) with a 2 × 2 design: correction approach (preemptive vs. responsive) × source (WHO vs. user) as compared to a control condition. Results suggest that the corrective infographic significantly improved public approval of the WHO's communication efforts and its credibility, but only when it was shared by a WHO information bot in direct response to a misinformation post. Interestingly, these same benefits did not accrue to the WHO when a social media user shared the WHO's graphics. Moreover, individuals initially critical of the WHO's performance increased their ratings of the WHO's credibility after seeing the corrective infographic from the WHO. The findings inform WHO's future RCCE efforts during a crisis and also highlight the potential of AI-driven bots affiliated with expert organizations to support corrective efforts.

It is broadly established that source cues matter for establishing information credibility (Pornpitakpan, 2004). Therefore, it is hardly surprising that when thinking about methods to address and correct misinformation, many scholars have pointed to the importance of source cues (Cook et al., 2017; Lewandowsky and van der Linden, 2021; Literat et al., 2021; Shen et al., 2019). But this research is plagued by inconsistent findings about which kinds of sources tend to be the most effective correctors (e.g., Berinsky, 2023; van der Meer and Jin, 2020; Vraga and Bode, 2017). We argue that research has not adequately accounted for recall of correction sources to better understand their effectiveness (but see Amazeen and Krishna, 2022; Swire-Thompson et al., 2023). For those reasons, in this project we use five separate experimental comparisons across three data collections to explore whether people can remember who offered a misinformation correction that they were exposed to, and whether source recall meaningfully affects the ability of a correction to increase accuracy. In general, we find relatively few people (38% to 54%) recall the source of a correction, speaking to the difficulties in making sources memorable. But we also demonstrate the importance of such recall: remembering the correction source in addition to the message itself boosts the impact of the correction on accuracy substantially. We reflect on the research needed to better understand how to boost source memorability and its implications for correction research.

Misinformation remains one of the most significant issues in the digital age. While automated fact-checking has emerged as a viable solution, most current systems are limited to evaluating factual accuracy. However, the detrimental effect of misinformation transcends simple falsehoods; it takes advantage of how individuals perceive, interpret, and emotionally react to information. This underscores the need to move beyond factuality and adopt more human-centered detection frameworks. In this survey, we explore the evolving interplay between traditional fact-checking approaches and psychological concepts such as cognitive biases, social dynamics, and emotional responses. By analyzing state-of-the-art misinformation detection systems through the lens of human psychology and behavior, we reveal critical limitations of current methods and identify opportunities for improvement. Additionally, we outline future research directions aimed at creating more robust and adaptive frameworks, such as neuro-behavioural models that integrate technological factors with the complexities of human cognition and social influence. These approaches offer promising pathways to more effectively detect and mitigate the societal harms of misinformation.

This study investigates whether source expertise (expert vs. non-expert), use of artificial intelligence (AI; AI vs. non-AI), and the placement (debunking vs. prebunking) of a correction influence its effectiveness in reducing misperceptions and intentions to consume raw milk. Results of a pre-registered two-wave online experiment ( N ₁ = 1,785, N ₂ = 1,568) suggest that debunking consistently reduces misperceptions and behavioral intentions for at least 1 week, while prebunking was less effective. Expert corrections only outperform non-expert corrections in reducing misperceptions in wave 1. In general, AI cues do not significantly influence the effectiveness of a correction, offering both opportunities and challenges for organizations hoping to automate corrections.

Of the many solutions to address political misinformation spreading on social media, user correction holds special promise for connective democracy given its emphasis on prioritizing user autonomy and fostering communication and connections across lines of disagreement. But for the connective democratic benefits to be realized, these user corrections should ideally come from those who express strong support for democratic norms. Using a nationally representative survey of Americans immediately after the 2020 U.S. presidential election, we find the opposite is true: self-reported correctors also tended to support political violence to achieve their goals. Rather than treating self-reported correction as a clear positive force for democracy, researchers and practitioners should consider the potential drawbacks and limitations of self-reported correction, particularly when coming from those with less supportive attitudes toward connective democracy.

The mixing of misinformation with high-quality news and information on social media has reinvigorated interest in the value of news literacy (NL) to build audience resiliency to misinformation. Optimizing NL messages for social media environments—where they may be seen alongside misinformation—allows these messages to reach audiences when they are most likely to benefit from them. Using a 2 (NL video vs. control video) x 2 (sunscreen promotion video vs. sunscreen misinformation video) online survey experiment (N = 780), we examine whether exposure to an NL video improves perceived personal NL skills and value for news literacy, as well as enables participants to recognize and avoid engaging with misinformation. Our findings suggest that after watching the NL video, individuals valued NL more but their self-perceived news literacy did not improve. Furthermore, watching the NL video made individuals rate the second video as less credible and reduced engagement with it no matter whether the second video contained misinformation or quality information. This research has several important implications. While watching an NL video could protect individuals by discrediting and decreasing engagement with misinformation, it may do so at the expense of high-quality information. We discuss the difficulty in designing NL messages that lead people to be appropriately skeptical and able to discern between high- and low-quality health information, rather than cynically disengaging with media content altogether.