đŸ€– Must-Read Articles đŸ€–

Experimental Feature

Even just looking backwards a week, there are a lot more articles published than most of us could hope to read. We can always skim the titles and abstracts ourselves, but I wanted to test out some automation. The articles below, all of which can be found elsewhere on this site among other new publications, were chosen by Google Gemini Flash Thinking as "must-read" articles. The proper criteria are in the eyes of the beholder and Gemini doesn't apply my criteria without error. I may continue to refine the prompting based on experience and feedback. Like the rest of the site, this will update daily!

DIMS: a no-code dashboard linking multimodal dynamic signals to communication practices
Communication Methods and Measures
Grace Qiyuan Miao, MichaƂ Weiss, Yuqian Ni, Jingjing Han
Full text
Centralized hashtags, embedded discourse: Two structural logics of youth climate activism on social media
European Journal of Communication
Yuzhe Zhu, Gustavo Cardoso
Full text
Youth-led climate action increasingly unfolds on digital platforms, where hashtags structurally connect issues, actors, and publics. Using hashtag co-occurrence data (2022–2024) from prominent youth climate activists on X and Portuguese youth climate activists on Instagram, this study applies link community analysis, which captures overlapping multi-membership structures that node-based clustering obscures. Two contrasting structural logics emerge. The international network exhibits a centralized configuration in which core climate hashtags such as #FridaysForFuture and #Peoplenotprofit occupy bridging positions across issue communities, so that visibility and structural centrality closely align. The Portuguese network displays a more horizontal configuration in which climate-related hashtags co-occur with feminism, education, domestic politics, and art-based activism, embedding climate discourse within everyday publics. Digital climate publics take shape as relational structures produced through patterns of hashtag overlap and cross-issue connectivity.
Real versus AI-generated war videos: comparative effects on viewer attitudes, emotional arousal, and behavioral intentions
Journal of Information Technology & Politics
Ravi Kumar
Full text
What Makes Child Harm Newsworthy? Event Characteristics, Victim Characteristics, and Media Attention
Journalism & Mass Communication Quarterly
Nikolay Ternov, Kuralay Mukhambetova, Zeinep Abetova
Full text
Media coverage shapes public understanding of social problems, yet harmful events receive uneven attention even when they involve children. This study examines factors shaping media visibility of child harm in Kazakhstan. Using a corpus of more than 230,000 Telegram news posts (2020–2025), we identify 1,571 incidents involving harm to children and estimate negative binomial regression models to test explanations derived from news values and ideal victim theory. Results show that event characteristics predict media attention more strongly than victim characteristics. Fatal interpersonal violence receives more coverage, while systemic failures receive less. Victim age and gender are not significant predictors.
Love Without Borders: Explaining Why People Interact with National Out-Groups on Tinder
Media Psychology
Ainhoa Arranz Aldana, Giulia Ranzini, Philipp K. Masur
Full text
Individual Political Attitudes and Group Norms Interact to Explain Biased Political Information Sharing
Media Psychology
Xinyao Zhang, Christin Scholz, Gijs Schumacher, Marte Otten
Full text
Where the World Connects: How Intergroup Contact on Social Media Influences Intergroup Outcomes
Media Psychology
Yansheng Liu, Yilan Guo, Alex Paloma, Mary Beth Oliver
Full text
Stress and self-efficacy under artificial intelligence disruption: A mixed-methods study on the usage of artificial intelligence among Finnish workers
New Media & Society
Atte Oksanen, Teo Keipi, Moona Heiskari, Iina Savolainen
Full text
Rapid workplace integration of artificial intelligence (AI) is reshaping work practices and challenging employees’ self-efficacy and well-being. This mixed-methods study examines these experiences using national survey data ( N = 2109) and qualitative interviews ( N = 54) with Finnish workers. Frequent AI use, AI-induced identity threat, and organizational AI deployment are associated with higher AI strain, whereas AI self-efficacy relates to frequent use, positive AI attitudes, and subjective well-being. Identity threat moderated several of these associations, highlighting the role of professional identity in AI-related experiences. Analysis drawing on Greimas’ semiotic modalities (must, can, know, and want) explains how these patterns are interpreted and negotiated in practice, revealing tensions between perceived self-efficacy and pressures to integrate AI. AI use was often described as voluntary. However, participants anticipated increasing competence demands and responsibility. The findings emphasize the need for organizational strategies that address identity-related challenges and support employee well-being in AI-driven work transformations.
Guardians of the cyberverse: Dispositional, attitudinal, and autobiographical correlates of user engagement against misogynist digital hate
New Media & Society
Rinat Meerson, Kevin Koban, Jörg Matthes
Full text
As digital misogyny remains widespread, user intervention is an essential complement to lenient platform governance. Yet, willingness to intervene varies considerably across individuals and cultural contexts. In a cross-national survey conducted in Austria ( n = 791), France ( n = 800), Hungary ( n = 775), and Sweden ( n = 785), we examine how dispositional, attitudinal, and autobiographical characteristics relate to users’ direct and indirect intervention and support for moderation. Empathic anger consistently emerged as the strongest positive predictor of all three outcomes across countries. Perspective-taking, online self-presentation, and victimization experiences showed meaningful positive associations with intervention, while misogynist attitudes were inconsistently related. Toxic disinhibition showed a dual pattern, negatively associated with moderation support but positively associated with direct intervention across all countries, suggesting its potential to be cleverly repurposed toward counterspeech. These findings underscore the importance of emotionally activating literacy programs and highlight the need for culturally sensitive intervention models.
Are Chinese scientists willing to engage with the public on genetically modified foods?
Public Understanding of Science
Yixuan Guo, Marcia D. de Barcellos, Lawton L. Nalley, Hans De Steur
Full text
Drawing on survey data from 321 Chinese scientists, this study investigates their willingness to engage on a controversial topic, namely, genetically modified foods. Extending the Theory of Planned Behavior with topic-specific beliefs, we analyze how six key factors shape engagement intentions: attitudes toward public engagement, perceived institutional and peer climates, perceived self-efficacy, scientists’ own attitude toward genetically modified foods, and their perception of public attitudes toward genetically modified foods. Results of partial least squares structural equation modeling reveal that attitude toward public engagement and perceived self-efficacy are the strongest predictors, while subjective norms play a limited role. Crucially, scientists’ own favorable attitudes toward genetically modified foods exert an indirect, positive effect on their engagement intentions. We further examine disciplinary differences in public engagement beliefs and find that agricultural and biological scientists hold significantly more positive attitudes toward engaging the public on genetically modified foods than scientists in other disciplines.
Social Media Platform Repertoires in Everyday Life: Measuring Social Media Use With Passive Metering
Social Media + Society
Moreno Mancosu, Mario Quaranta, Fabio Torreggiani
Full text
Social media platforms structure the ways in which billions of citizens access information. Nevertheless, to assess social media time use, most research relies on limited data sources, such as self-reported survey measures or behavioral data from single platforms or convenience samples. Using 6 weeks of passive metering of smartphone and computer activity from a quota sample of Italian adults (18–70, n = 1,457), this study provides a behaviorally measured account of platform-specific social media use across the adult population. We quantify time allocation across Facebook, Instagram, TikTok, YouTube, and Twitter/X and examine how use varies by sociodemographic factors. Results show that Facebook and YouTube dominate everyday social media time, accounting for nearly three-quarters of consumption, while Instagram attracts comparable average time from a narrower user base, and TikTok shows concentrated usage among younger adopters. Twitter/X occupies a marginal niche. Use is overwhelmingly app-based (>90%), a pattern that browser-based measurement cannot capture. Age is the strongest demographic predictor: Gen Z concentrates attention on Instagram and TikTok, while older cohorts favor Facebook; YouTube shows weak demographic patterning. Gender and education are consistent secondary predictors: women devote more time to Facebook, Instagram, and social media overall, and higher education is associated with lower aggregate use and a smaller share on Facebook and TikTok. Employment and income show null effects. Together, these behavioral, cross-platform figures provide a more accurate basis for weighting platforms by their actual audiences than self-reported or single-platform evidence allows.
Automated Influence: How Social Bots Reshape Discursive Opportunities in Online Activism
Social Media + Society
Linda Li, Helen Margetts
Full text
Social bots are an increasingly prominent presence in online protest discourse. We adapt the discursive opportunity structure (DOS) framework for digital protest, refining its dimensions of visibility, resonance and legitimacy to account for platform-mediated discourse and automated activity. Applying the refined framework through computer-assisted content analysis of bot activity in two major protests – Black Lives Matter (BLM) and Extinction Rebellion (XR) – we move beyond treating bots as a monolithic or uniformly disruptive force, identifying competing pro- and counter-protest strategies distributed differently across the three dimensions. In BLM, pro-protest bot content concentrates on cultural resonance, whereas counter-bot content targets legitimacy; in XR, pro-protest bots are used to build legitimacy and participatory visibility, whereas counter-bots more often work to downplay the movement’s salience. The framework’s granularity discriminates between movement types in ways coarser accounts of ‘political bots’ cannot, showing that understanding contemporary movements requires attending to how automated activity is deployed within their discourse.
Digital turnout mobilization: Enhancing voter turnout with digital messaging
European Journal of Political Research
Kasper M. Hansen, Esther EjrnĂŠs Bastholm
Full text
Voter turnout is declining across Europe, raising concerns about democratic legitimacy and prompting a search for effective mobilization strategies. The new digital era has transformed the landscape for political communication, making it possible to mobilize voters through direct digital messages. These messages have proven to be a low-cost and easily scalable tool, capable of reaching millions of voters efficiently and with broad coverage. We synthesize existing evidence through a comprehensive meta-analysis of voter mobilization experiments from the past decade that apply digital messages as a Get-Out-The-Vote (GOTV) tool. The analysis assesses the overall effectiveness of digital messaging in increasing turnout, with particular attention to its potential for engaging marginalized groups and promoting participatory equality. Second, we present results from a new population-wide GOTV experiment conducted during the 2024 European Parliament election in Denmark – the first to apply digital messages to the entire electorate. The intervention increased turnout by 0.5 percentage points. By using year-of-birth bins to estimate effects across all age cohorts and disaggregating the immigrant category into Western and non-Western subpopulations, we can examine previously unexplored in-depth heterogeneous effects. We find that digital messages consistently have positive effects on turnout, including among marginalized voters.
Perceiving Politics: Considering the Political Categorization of Faces
Journal of Experimental Political Science
Thomas Bergeron, Blake Lee-Withing, Natasha Goel
Full text
Categorizing people into groups is a fundamentally human trait, but we know very little about how people assign political identities to strangers. Across two innovative survey experiments, we consider how people politically categorize others according to a single source of information: faces. We find that displays of emotions, particularly negative emotions, predict the political sorting of strangers. Faces featuring negative emotions are categorized as out-partisans, whereas faces exhibiting happiness are categorized as co-partisans.
The Synthetic Imputation Approach: Using LLMs to Generate Synthetic Texts for Underrepresented Categories in Supervised Classification Tasks
Political Analysis
Joan C. Timoneda
Full text
Annotated text data are central to many empirical applications in political science, yet many annotation tasks involve categories that are sparsely populated. These low- N categories pose problems for both encoder-only and decoder-only (generative) Transformer large language models (LLMs), leading to low performance, biased evaluation and measurement error in downstream analyses. This article introduces the synthetic imputation approach, a targeted method for expanding sparse categories while remaining consistent with the category-defining content of the human-coded examples. The approach uses a generative LLM to produce additional texts for underrepresented categories by conditioning on small, rotating subsets of human-coded examples. Rather than paraphrasing or replacing the original data, synthetic imputation adds informative variation only where data are thin. I evaluate the approach using both encoder-only and decoder-only LLM classifiers across multiple political science applications. The results show that synthetic imputation substantially improves performance and evaluation stability in sparse settings, recovering full-sample performance once a modest number of true observations is available, with limited and predictable overfitting at very low levels of N .
Do Appeals to the Diaspora Affect their Attachment to the Homeland? Findings from Jewish-Americans before and during the Israel-Hamas War
Political Behavior
Yehonatan Abramson, Anil Menon, Alon Yakter
Full text
Due to the growing importance of diasporas in contemporary geopolitics, many states seek to cultivate diasporic attachment and encourage diaspora mobilization using strategic narratives. Are such appeals effective? We report results from a pre-registered two-wave survey experiment conducted among Jewish Americans before and during the Israel-Hamas war. Our intervention assesses the impact of three common appeals identified by prior work: threats to the homeland, shared historical memory, and discrimination in the diaspora’s country of residence. We find that none of these appeals influenced Jewish-Americans’ attachment to, support for, or intent to mobilize for Israel, even during a war in the “homeland.” Instead, baseline differences in these outcomes reflect deep-seated generational, denominational, and partisan cleavages within the Jewish American community. These findings suggest that intra-diaspora fragmentation and long-standing views of the “homeland” might dominate short-term rhetorical interventions. Consequently, strategic frames deployed by kin states and political animators may be less effective on diaspora audiences than intended.
Relational Clientelism, Information Processing, and Accountability
Political Behavior
Miquel Pellicer, Eva Wegner
Full text
Research shows that providing information on incumbent programmatic performance often has limited impact on citizen attitudes towards incumbents, even if citizens tend to care strongly about corruption and service delivery. We use a survey experiment in South Africa and Tunisia to study if relational clientelism, a type of clientelism often valued by clients, plays a role in this puzzle. We present respondents with a radiocast that varies the profile of the politician (clientelistic vs. neutral) and his performance in service delivery and corruption (good vs. bad). We find that, when clientelistic politicians are valued, performance information on them matters less than for neutral politicians. The effect appears to operate via motivated reasoning: Respondents are less likely to process negative performance information when exposed to the clientelistic profile. Our findings suggest that, in certain contexts, relational clientelism is a valued candidate characteristic that can offset bad programmatic performance akin to co-partisanship or co-ethnicity.
Egotropic and sociotropic negativity bias in social policy reform preferences
Political Psychology
Tijs Laenen, Elise Aerts
Full text
Although negativity bias is widely documented in political psychology, evidence is far more mixed when it comes to voters' preferences for social policy reform, particularly when they are confronted with trade‐offs between personal and societal outcomes. Building on existing theories of political communication, this article examines how voters evaluate policy reforms when considering realistic information about potential losses and gains. Empirically, we analyze a simu‐survey experiment examining how voters update their reform preferences in response to information derived from microsimulation modeling about the predicted effects of policy reform on their pocketbook and on the poverty level. The results show that negative information weighs more heavily than positive information in shaping reform preferences at both the egotropic and sociotropic levels. In the case of competing information, the effects are asymmetric: Positive pocketbook information tends to be neutralized by negative poverty information, whereas negative pocketbook information outweighs positive poverty information. However, egotropic negativity bias is most pronounced for smaller pocketbook losses, which are more likely to violate voters' prior expectations than larger losses. Consistent with this pattern, expectation violations—rather than differences in recall of, or trust in, information—seems the primary mechanism underlying negativity bias in policy preferences.
Polling Without People: Can Silicon Samples of CASM-Inspired LLM Personae Mimic Human Respondents in Pre-Election Polling?
PS: Political Science & Politics
Danielius Jonaitis, Brendan Corcoran, Binam Poudyal, Alexandra Roskam, Ava Alan Ali, Shriya Agrawal, Raymond Duch, Roberto Cerina
Full text
Survey Behavior and Cognition Across the Life Course: Evidence from the Panel Study of Income Dynamics
Journal of Survey Statistics and Methodology
Malak B Kalasho, Esther M Friedman, Thomas F Crossley
Full text
Cognitive abilities across adulthood are important pre-cursors to dementia. The primary way to capture cognitive abilities is via expensive and time-consuming standardized cognitive tests. We explore the extent to which cognition can be inferred from indicators of survey response behavior (SRB), say, the time to complete a questionnaire and the incidence of contradictory responses. We use nationally representative data from a mixed-mode supplement to the Panel Study of Income Dynamics collected in 2016. This Wellbeing and Daily Life supplement includes rich cognitive data capturing both verbal and quantitative reasoning, and paradata from the supplement (e.g., time to complete). We assess the ability of survey response behavior to improve prediction of cognitive ability after accounting for standard sociodemographic variables typically available in surveys. We explore this from age 30 and across the life course. We find that SRB measures provide modest but consistent explanatory power across a wide range of ages. The explanatory power of SRB measures in our data is less than that of education, except at the oldest ages. SRB measures contain useful additional information, but models that include basic demographics, education and SRB measures leave much of the variation in cognitive ability unexplained. Which SRB measures best predict cognitive ability is consistent across cognitive measure but differs by age. Errors (i.e., contradictory responses) and response times are the best predictors at working ages but skips and backups are more predictive at older ages.
Restructuring Scientific Papers for Human and AI Readers
Advances in Methods and Practices in Psychological Science
Zhicheng Lin
Full text
Scientific communication faces a dual crisis: Exponential publication growth—now accelerated by writing assisted by artificial intelligence (AI)—overwhelms human readers and reviewers, while fragmented evidence impedes automated synthesis. This fragmentation is especially acute in the behavioral and social sciences, where poorly standardized stimulus sets and jingle-jangle measurement fallacies (identical labels for distinct constructs; different labels for the same construct) hinder comparison, while underreported context conceals effect heterogeneity. Current AI tools can summarize papers but cannot reliably synthesize findings across incommensurable studies; they also risk perpetuating biases when trained on unstructured, unverified text. I propose restructuring scientific papers for two audiences: front-loaded narratives for time-pressed human readers, paired with research-object packages containing executable environments, semantic annotations, and tidy trial-level data. This design makes papers queryable: Readers can interrogate data and probe analytic choices, while the packages ground automated verification and AI-assisted manuscript evaluation—before submission or during peer review—in inspectable data and executable analyses rather than narrative claims alone. Such papers could become nodes in continuously updated evidence networks, automatically contributing effect sizes to living, versioned meta-analyses and allowing corrections and retractions to propagate through dependent syntheses. Widespread adoption will require institutions to credit research-object packages as scholarly contributions and invest in the open standards, infrastructure, training, and governance needed for their creation and long-term stewardship.
Introducing GABLE: Lightweight infrastructure for longitudinal experiments using Google Apps Script
Behavior Research Methods
Ibrahim Berber, Ilke Kas, Tarun Sepuri, Brooke N. Macnamara, M. Cenk ÇavuƟoğlu, Amy L. Wilson-Delfosse, Elizabeth A. Krupinski, Philip J. Smith, Soumya Ray
Full text
Longitudinal studies are essential for understanding how behaviors, skills, and psychological states evolve over time. While online platforms have made study development easier, researchers still face major logistical challenges when managing multi-session studies, especially when coordinating large numbers of participants, session timing, reminders, and compensation. In this paper, we introduce GABLE ( G oogle A pps script- B ased L ongitudinal E xperimentation), a lightweight infrastructure built using Google Apps Script. GABLE automates participant scheduling, e-mail communication, session tracking, and compensation. We demonstrate its effectiveness through two real-world psychological studies involving more than 850 participants with 20 sessions, and discuss its benefits, limitations, and potential extensions. GABLE addresses many challenges of running large online longitudinal studies while limiting the need for custom infrastructure or manual intervention.
Human-in-the-loop LLM-assisted thematic analysis: A validated, reproducible tutorial for large-scale qualitative research
Behavior Research Methods
Mandana Jafarian
Full text
Estimating the threat of AI-agent responding across online survey platforms
Proceedings of the National Academy of Sciences
Grace Zhang, Robert Walatka, Stephanie Y. Chen, Alexander K. Moore, Oleg Urminsky, Kianté Fernandez, Andrea Low, Jonathan E. Bogard, Craig R. Fox
Full text
Recent research and advances in LLMs have led to widespread concern that AI agents could pose as human online survey-takers. However, it remains unclear how prevalent AI agents are on survey research platforms and how to effectively detect AI agents. We validated a series of AI detection tests that effectively separated verified-human participants from both prompted commercial and purpose-built AI agents. Using these tests, we collected surveys on seven online platforms and find wide variation in rates of respondents failing AI tests, ranging from 6 to 41% across platforms (compared to a 2.4% false-positive rate for in-person human participants). We further demonstrate that undetected AI agents can impact the results of online surveys. Our findings suggest that while there is an urgent need for AI detection tests and consistent, systematic monitoring of data quality on online platforms, some platforms currently appear to provide data with a low rate of AI agents.
AI can persuade people to take political actions
Proceedings of the National Academy of Sciences
Kobi Hackenburg, Luke Hewitt, Caroline Wagner, Christopher Summerfield, Ben M. Tappin
Full text
There is substantial concern about the ability of advanced AI to influence people’s behavior. A rapidly growing body of research has found that AI can produce large persuasive effects on people’s attitudes, but whether AI can persuade people to take consequential real-world actions has remained unclear. In two large preregistered experiments ( N = 17 , 950 responses from 14,779 people), we used conversational AI models to persuade participants on a range of attitudinal and behavioral outcomes, including signing real petitions and donating money to charity. We found sizable AI persuasion effects on these behavioral outcomes (e.g. + 19.7 percentage points on petition signing). However, we observed no evidence of a correlation between AI persuasion effects on attitudes and effects on behavior. Moreover, we replicated prior findings that information provision drove effects on attitudes, but found no such evidence for our behavioral outcomes. In a test of eight behavioral persuasion strategies, the strategy focused on information provision produced the smallest behavioral effect, though differences among the eight were small. Taken together, these results suggest that previous findings relying on attitudinal outcomes may generalize poorly to behavior, and therefore risk substantially mischaracterizing the real-world behavioral impact of AI persuasion.