Ádám Stefkovics
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Ádám Stefkovics has a total of 34 co-authors. Here are the top co-authors by number of shared publications:
- Júlia Koltai
MTA–TK Lendület “Momentum” Digital Social Science Research Group for Social Stratification, ELTE Centre for Social Sciences, Budapest, Hungary
3 shared publications - Zoltán Kmetty
CSS Research Group, Centre for Social Sciences
3 shared publications - Anna Sára Ligeti
ELTE Centre for Social Sciences, National Laboratory for Health Security , Budapest
2 shared publications Blanka Zsófia Grubits
Századvég Foundation, Budapest, Hungary
2 shared publications- Cindy Xiong Bearfield
Interactive Computing, Georgia Tech, Atlanta, Georgia, USA
2 shared publications - Dömötör Gere
Szazadveg Foundation , Political Analysis Research Institute, 1037 Budapest
2 shared publications
Latest Publications
- Open Access
- 15 September 2026
- Published by Elsevier in Computers in Human Behavior Reports
- Vol. 24
Determinants of AI acceptance in the workplace: Evidence from eleven European countries
- 15 September 2026
- Published by Elsevier in Computers in Human Behavior Reports
- Vol. 24
Open AccessDeterminants of AI acceptance in the workplace: Evidence from eleven European countries
- Open Access
- 5 July 2026
- Published by Akademiai Kiado Zrt. in Orvosi Hetilap
- Vol. 167 (27) , 1059-1067
A Védőoltási Attitűdök Vizsgálata (Vaccination Attitudes Examination – VAX) skála magyar nyelvű validálása
Bevezetés: Az oltásokkal kapcsolatos attitűdök fontos szerepet játszanak az oltási magatartás alakításában, ugyanakkor Magyarországon jelenleg nem áll rendelkezésre validált, általános felnőtt populációra alkalmazható mérőeszköz ezen attitűdök pszichometriailag megalapozott vizsgálatára. Célkitűzés: A kutatás célja a Védőoltási Attitűdök Vizsgálata (Vaccination Attitudes Examination – VAX) skála magyar nyelvű adaptációja és pszichometriai validálása a felnőtt magyar lakosság körében. Módszer: A vizsgálat országosan reprezentatív, 18 éves és idősebb felnőttekből álló mintán (n = 1000) végzett keresztmetszeti adatfelvétel adatai alapján zajlott. A skála faktorstruktúráját feltáró és megerősítő faktoranalízissel vizsgáltuk (az előbbit „maximum likelihood” becsléssel és oblimin rotációval, az utóbbit ordinális tételekre alkalmas DWLS-becsléssel). A belső konzisztenciát Cronbach-alfa, valamint CFA-alapú megbízhatósági mutatók segítségével értékeltük. A konvergens validitást elméletileg releváns attitűd- és bizalmi változókkal való összefüggések elemzésével teszteltük. Eredmények: Az eredmények egy elméletileg koherens, négydimenziós faktorstruktúrát támasztottak alá. A skála és alskálái jó vagy kiváló belső megbízhatóságot mutattak (α = 0,69–0,92). A VAX-pontszámok erős negatív összefüggést jeleztek az összeesküvés-elméletekbe vetett hittel, valamint mérsékelt-erős negatív kapcsolatot a tudományba és az intézményekbe vetett bizalommal. A COVID–19 ellen oltatlan válaszadók szignifikánsan magasabb pontszámokat értek el valamennyi alskálán. Megbeszélés: A magyar nyelvű VAX-skála megbízhatóan és érvényesen méri az oltásokkal kapcsolatos szkeptikus attitűdök többdimenziós szerkezetét. Az eredmények összhangban állnak a nemzetközi validációs tanulmányokkal, és alátámasztják a skála alkalmazhatóságát a magyar kontextusban. Következtetés: A VAX-skála magyar változata alkalmas eszköz az oltási attitűdök kutatására, valamint az oltási bizonytalanság társadalmi és pszichológiai hátterének feltárására. Orv Hetil. 2026; 167(27): 1059–1067.
- 5 July 2026
- Published by Akademiai Kiado Zrt. in Orvosi Hetilap
- Vol. 167 (27) , 1059-1067
Open AccessA Védőoltási Attitűdök Vizsgálata (Vaccination Attitudes Examination – VAX) skála magyar nyelvű validálása
Bevezetés: Az oltásokkal kapcsolatos attitűdök fontos szerepet játszanak az oltási magatartás alakításában, ugyanakkor Magyarországon jelenleg nem áll rendelkezésre validált, általános felnőtt populációra alkalmazható mérőeszköz ezen attitűdök pszichometriailag megalapozott vizsgálatára. Célkitűzés: A kutatás célja a Védőoltási Attitűdök Vizsgálata (Vaccination Attitudes Examination – VAX) skála magyar nyelvű adaptációja és pszichometriai validálása a felnőtt magyar lakosság körében. Módszer: A vizsgálat országosan reprezentatív, 18 éves és idősebb felnőttekből álló mintán (n = 1000) végzett keresztmetszeti adatfelvétel adatai alapján zajlott. A skála faktorstruktúráját feltáró és megerősítő faktoranalízissel vizsgáltuk (az előbbit „maximum likelihood” becsléssel és oblimin rotációval, az utóbbit ordinális tételekre alkalmas DWLS-becsléssel). A belső konzisztenciát Cronbach-alfa, valamint CFA-alapú megbízhatósági mutatók segítségével értékeltük. A konvergens validitást elméletileg releváns attitűd- és bizalmi változókkal való összefüggések elemzésével teszteltük. Eredmények: Az eredmények egy elméletileg koherens, négydimenziós faktorstruktúrát támasztottak alá. A skála és alskálái jó vagy kiváló belső megbízhatóságot mutattak (α = 0,69–0,92). A VAX-pontszámok erős negatív összefüggést jeleztek az összeesküvés-elméletekbe vetett hittel, valamint mérsékelt-erős negatív kapcsolatot a tudományba és az intézményekbe vetett bizalommal. A COVID–19 ellen oltatlan válaszadók szignifikánsan magasabb pontszámokat értek el valamennyi alskálán. Megbeszélés: A magyar nyelvű VAX-skála megbízhatóan és érvényesen méri az oltásokkal kapcsolatos szkeptikus attitűdök többdimenziós szerkezetét. Az eredmények összhangban állnak a nemzetközi validációs tanulmányokkal, és alátámasztják a skála alkalmazhatóságát a magyar kontextusban. Következtetés: A VAX-skála magyar változata alkalmas eszköz az oltási attitűdök kutatására, valamint az oltási bizonytalanság társadalmi és pszichológiai hátterének feltárására. Orv Hetil. 2026; 167(27): 1059–1067.
- Open Access
- 9 May 2026
- Published by Taylor & Francis in International Journal of Social Research Methodology
- p. 1-21
How do people answer web surveys? The consequences of distractions, multitasking, and completion context
- 9 May 2026
- Published by Taylor & Francis in International Journal of Social Research Methodology
- p. 1-21
Open AccessHow do people answer web surveys? The consequences of distractions, multitasking, and completion context
- 22 April 2026
- Published by Oxford University Press (OUP) in Public Opinion Quarterly
Gender-of-Interviewer Effects in the Measurement of Gender Attitudes: Lessons from 18 Countries of the European Social Survey
Measuring gender-related questions in face-to-face surveys can be challenging due to the potential influence of the interviewer’s gender. First, people may feel more comfortable disclosing sensitive information (e.g., gender-based discrimination) to a female interviewer or an interviewer of the same gender. Second, respondents may alter their responses in light of what they perceive as desirable, for instance, by reporting more feminist views when interviewed by a female interviewer. To test these assumptions, we use the gender module of the eleventh round of the European Social Survey. We analyze patterns from 28,567 respondent-interviewer dyads from 18 countries. We find no evidence of the interviewer’s gender influencing disclosure and limited evidence of gender-of-interviewer (GOI) effects on distributions. These findings are consistent across countries with varying levels of gender equality. Our results indicate that in European societies, GOI effects may not cause significant harm in social surveys.
- 22 April 2026
- Published by Oxford University Press (OUP) in Public Opinion Quarterly
Gender-of-Interviewer Effects in the Measurement of Gender Attitudes: Lessons from 18 Countries of the European Social Survey
Measuring gender-related questions in face-to-face surveys can be challenging due to the potential influence of the interviewer’s gender. First, people may feel more comfortable disclosing sensitive information (e.g., gender-based discrimination) to a female interviewer or an interviewer of the same gender. Second, respondents may alter their responses in light of what they perceive as desirable, for instance, by reporting more feminist views when interviewed by a female interviewer. To test these assumptions, we use the gender module of the eleventh round of the European Social Survey. We analyze patterns from 28,567 respondent-interviewer dyads from 18 countries. We find no evidence of the interviewer’s gender influencing disclosure and limited evidence of gender-of-interviewer (GOI) effects on distributions. These findings are consistent across countries with varying levels of gender equality. Our results indicate that in European societies, GOI effects may not cause significant harm in social surveys.
- Open Access
- 23 February 2026
- Published by Oxford University Press (OUP) in PNAS Nexus
- Vol. 5 (3)
Beliefs and sharing intentions of human- and AI-generated fake news: Evidence from 27 European countries
Misinformation remains a major challenge in today’s information environment, and rapid advances in AI-driven content generation risk amplifying this problem. Generative AI represents a double-edged sword: beyond its growing utility for detecting misinformation, it can also facilitate democratic deliberation, counter conspiracy narratives, and promote reliable information, even as the same technologies enable the rapid, large-scale production of persuasive false content. Understanding how people perceive AI-generated misinformation is therefore crucial for designing effective interventions and safeguarding information integrity. To address this, we embedded a preregistered experiment in a large-scale web survey conducted across 27 European countries. Participants were presented with eight short news headlines related to the Russo-Ukrainian war: four AI-generated and four human-generated, evenly split between real and fake news. For each headline, respondents assessed its perceived veracity and their willingness to share it. Our findings show that fake news is consistently viewed as less accurate and less likely to be shared, with systematic differences across countries and individual characteristics such as cognitive reflection, ideology, and trust. While differences between human- and AI-generated content were minimal, the results reveal broader and robust patterns in how people evaluate misinformation across diverse European contexts. These insights highlight the need to strengthen individuals’ cognitive and informational resilience to counter the spread of misleading content in increasingly complex media environments.
- 23 February 2026
- Published by Oxford University Press (OUP) in PNAS Nexus
- Vol. 5 (3)
Open AccessBeliefs and sharing intentions of human- and AI-generated fake news: Evidence from 27 European countries
Misinformation remains a major challenge in today’s information environment, and rapid advances in AI-driven content generation risk amplifying this problem. Generative AI represents a double-edged sword: beyond its growing utility for detecting misinformation, it can also facilitate democratic deliberation, counter conspiracy narratives, and promote reliable information, even as the same technologies enable the rapid, large-scale production of persuasive false content. Understanding how people perceive AI-generated misinformation is therefore crucial for designing effective interventions and safeguarding information integrity. To address this, we embedded a preregistered experiment in a large-scale web survey conducted across 27 European countries. Participants were presented with eight short news headlines related to the Russo-Ukrainian war: four AI-generated and four human-generated, evenly split between real and fake news. For each headline, respondents assessed its perceived veracity and their willingness to share it. Our findings show that fake news is consistently viewed as less accurate and less likely to be shared, with systematic differences across countries and individual characteristics such as cognitive reflection, ideology, and trust. While differences between human- and AI-generated content were minimal, the results reveal broader and robust patterns in how people evaluate misinformation across diverse European contexts. These insights highlight the need to strengthen individuals’ cognitive and informational resilience to counter the spread of misleading content in increasingly complex media environments.
- Open Access
- 9 September 2025
- Published by SAGE Publications in Sociological Methods & Research
- Vol. 55 (4) , 1188-1221
What Types of Survey Questions are Prone to Interviewer Effects? Evidence Based on 29,000 Intra-Interviewer Correlations From 28 Countries of the European Social Survey
Interviewer effects are a common challenge in face-to-face surveys. Understanding the conditions that make interviewer variance more likely to occur is essential in tackling sources of bias. Earlier evidence suggests that certain features of the survey instrument provide more ground for interviewer influence. For instance, attitudinal, sensitive, complex or open-ended questions invite more interviewer variance. In this article, we aim to validate earlier results, previously derived from single-country studies, by using the large cross-national sample of the European Social Survey (ESS). We compare 29,330 intra-interviewer correlations derived from 984 survey questions from 28 countries using data from 10 waves of the ESS. The questions were manually coded based on several characteristics. These features of survey questions were then used as predictors of intraclass correlations (ICCs) in multilevel models. The results show that question characteristics account for a significant portion of the variation in ICCs, with certain types, such as attitude and non-factual questions, items appearing later in the survey, and those using showcards, being especially susceptible to interviewer effects. Our findings have important implications for both interviewer training and questionnaire design.
- 9 September 2025
- Published by SAGE Publications in Sociological Methods & Research
- Vol. 55 (4) , 1188-1221
Open AccessWhat Types of Survey Questions are Prone to Interviewer Effects? Evidence Based on 29,000 Intra-Interviewer Correlations From 28 Countries of the European Social Survey
Interviewer effects are a common challenge in face-to-face surveys. Understanding the conditions that make interviewer variance more likely to occur is essential in tackling sources of bias. Earlier evidence suggests that certain features of the survey instrument provide more ground for interviewer influence. For instance, attitudinal, sensitive, complex or open-ended questions invite more interviewer variance. In this article, we aim to validate earlier results, previously derived from single-country studies, by using the large cross-national sample of the European Social Survey (ESS). We compare 29,330 intra-interviewer correlations derived from 984 survey questions from 28 countries using data from 10 waves of the ESS. The questions were manually coded based on several characteristics. These features of survey questions were then used as predictors of intraclass correlations (ICCs) in multilevel models. The results show that question characteristics account for a significant portion of the variation in ICCs, with certain types, such as attitude and non-factual questions, items appearing later in the survey, and those using showcards, being especially susceptible to interviewer effects. Our findings have important implications for both interviewer training and questionnaire design.
- Open Access
- 7 August 2025
- Published by Center for Open Science in SocArXiv
Beliefs and Sharing Intentions of Human- and AI-Generated Fake News: Evidence from 27 European Countries
Misinformation remains a major challenge in today’s information environment, and rapid advances in AI-driven content generation risk amplifying this problem. Generative AI represents a double-edged sword: it holds great promise for improving the detection of false information, yet it also enables the rapid, large-scale production of highly persuasive fake content. Understanding how people perceive AI-generated misinformation is therefore crucial for designing effective interventions and safeguarding information integrity.To address this, we embedded a pre-registered experiment in a large-scale web survey conducted across 27 European countries. Participants were presented with eight short news headlines related to the Russo-Ukrainian war: four AI-generated and four human-generated, evenly split between real and fake news. For each headline, respondents assessed its perceived veracity and their willingness to share it.Our findings show that, regardless of authorship, fake news is less likely to be perceived as accurate and less likely to be shared, although differences between human- and AI-generated content were minimal. This pattern was remarkably consistent across the 27 countries, with some variation by individual characteristics. Importantly, our study demonstrates that GPT-4 models can generate convincing fake news on the Russo-Ukrainian war from a simple prompt, producing content perceived as equally – if not more – credible than human-written news.
- 7 August 2025
- Published by Center for Open Science in SocArXiv
Open AccessBeliefs and Sharing Intentions of Human- and AI-Generated Fake News: Evidence from 27 European Countries
Misinformation remains a major challenge in today’s information environment, and rapid advances in AI-driven content generation risk amplifying this problem. Generative AI represents a double-edged sword: it holds great promise for improving the detection of false information, yet it also enables the rapid, large-scale production of highly persuasive fake content. Understanding how people perceive AI-generated misinformation is therefore crucial for designing effective interventions and safeguarding information integrity.To address this, we embedded a pre-registered experiment in a large-scale web survey conducted across 27 European countries. Participants were presented with eight short news headlines related to the Russo-Ukrainian war: four AI-generated and four human-generated, evenly split between real and fake news. For each headline, respondents assessed its perceived veracity and their willingness to share it.Our findings show that, regardless of authorship, fake news is less likely to be perceived as accurate and less likely to be shared, although differences between human- and AI-generated content were minimal. This pattern was remarkably consistent across the 27 countries, with some variation by individual characteristics. Importantly, our study demonstrates that GPT-4 models can generate convincing fake news on the Russo-Ukrainian war from a simple prompt, producing content perceived as equally – if not more – credible than human-written news.
- Open Access
- 21 July 2025
- Published by Center for Open Science in OSF Preprints
The Accuracy of Survey Measures on Political Activity on Facebook. Evidence from a data donation study
Self-reported survey measures of social media use are commonly used across disciplines, yet the validity of findings in this area depends heavily on the accuracy of these reports. In political communication research, measurement error in reported online political behavior can distort estimates of media effects and lead to misleading conclusions about online political engagement. To address this concern, our study leveraged donated digital trace data from 758 Hungarian Facebook users and compared it with responses to standard survey questions provided by the same individuals. In addition, in a pre-registered experiment, we varied the ordering of the survey items. Our findings reveal moderate correlations between self-reported and actual digital behavior. General items showed closer alignment with observed behavior than self-reports of specific political activities. Participants typically underestimated their general Facebook activities and overestimated their political engagement on Facebook. Importantly, some of this discrepancy may stem not from respondents’ inaccuracy but from difficulties in classifying digital actions as politically relevant. Survey measures were more reliable in the general-specific questionnaire design and among individuals with high political interest. We discuss potential avenues for improving survey-based measures and emphasize the need for further research to address the methodological challenges associated with the use of digital trace data.
- 21 July 2025
- Published by Center for Open Science in OSF Preprints
Open AccessThe Accuracy of Survey Measures on Political Activity on Facebook. Evidence from a data donation study
Self-reported survey measures of social media use are commonly used across disciplines, yet the validity of findings in this area depends heavily on the accuracy of these reports. In political communication research, measurement error in reported online political behavior can distort estimates of media effects and lead to misleading conclusions about online political engagement. To address this concern, our study leveraged donated digital trace data from 758 Hungarian Facebook users and compared it with responses to standard survey questions provided by the same individuals. In addition, in a pre-registered experiment, we varied the ordering of the survey items. Our findings reveal moderate correlations between self-reported and actual digital behavior. General items showed closer alignment with observed behavior than self-reports of specific political activities. Participants typically underestimated their general Facebook activities and overestimated their political engagement on Facebook. Importantly, some of this discrepancy may stem not from respondents’ inaccuracy but from difficulties in classifying digital actions as politically relevant. Survey measures were more reliable in the general-specific questionnaire design and among individuals with high political interest. We discuss potential avenues for improving survey-based measures and emphasize the need for further research to address the methodological challenges associated with the use of digital trace data.
- 24 April 2025
- Published by Springer Nature in Comparative European Politics
- Vol. 23 (4) , 593-614
Who's left, who's right? Examining cross-national variations in the use of the left–right scale with anchoring vignettes
- 24 April 2025
- Published by Springer Nature in Comparative European Politics
- Vol. 23 (4) , 593-614
Who's left, who's right? Examining cross-national variations in the use of the left–right scale with anchoring vignettes