Conference Agenda
Overview and details of the sessions of this conference. Please select a date or location to show only sessions at that day or location. Please select a single session for detailed view (with abstracts and downloads if available).
|
Daily Overview |
| Session | ||
16 SES 02 A
Paper Session
| ||
    | ||
16. ICT in Education and Training
Paper Enjoyment, Satisfaction, Anxiety or Boredom? A Three-Level Meta-Analysis of Generative AI's Influence on Learners' Emotions Through the Lens of Control-Value Theory University of Cambridge, United Kingdom Presenting Author:Theme and Research Overview This research addresses learners' emotional responses to generative artificial intelligence in educational contexts, a critical yet underexplored dimension of AI-supported learning. While existing research has focused on cognitive outcomes, emotional implications of generative AI use remain fragmented and theoretically underdeveloped. This meta-analysis synthesizes evidence from 25 experimental and quasi-experimental studies, examining how generative AI influences four theoretically distinct emotion categories grounded in Control-Value Theory: positive-activating (e.g., enjoyment), positive-deactivating (e.g., satisfaction), negative-activating (e.g., anxiety), and negative-deactivating (e.g., boredom) emotions. The study employs rigorous three-level meta-analytic techniques to account for dependent effect sizes, yielding 51 effect sizes across diverse educational settings. Findings reveal differentiated emotional effects: significant increases in positive emotions and reductions in anxiety, alongside context-dependent effects on boredom moderated by intervention duration. Notably, emotional outcomes remain largely consistent across educational levels and disciplines, suggesting broad applicability of implementation principles. European Dimension This work holds particular significance for European education. As the European Higher Education Area integrates generative AI technologies, evidence-based guidance on their affective implications becomes essential. The research directly responds to European Commission priorities on responsible AI integration, providing insights applicable across diverse national contexts, pedagogical traditions, and regulatory frameworks within EU member states. By synthesizing international evidence, this study positions European research within global developments while delivering findings that support cross-national policy development and institutional AI adoption strategies across Europe. Research Questions This study investigates four interconnected research questions grounded in Control-Value Theory:
These questions address fundamental gaps in international educational technology research by examining affective dimensions of AI-mediated learning—an area of growing concern for European educators and policymakers. Theoretical Framework The study is rigorously grounded in Control-Value Theory (Pekrun, 2006, 2024), which explains how learners' emotions emerge from subjective appraisals of control over learning activities and the value attached to those activities. Different combinations of perceived control and task value give rise to qualitatively distinct emotions organized along two dimensions: valence (positive versus negative) and activation (activating versus deactivating). This framework yields four emotion categories: positive-activating (e.g., enjoyment, hope), positive-deactivating (e.g., satisfaction, relief), negative-activating (e.g., anxiety, frustration), and negative-deactivating (e.g., boredom, hopelessness). Control-Value Theory is particularly well suited to examining generative AI because these systems can substantially reshape learners' emotional appraisals. AI tools may enhance perceived control through immediate feedback, adaptive explanations, and personalized scaffolding, but may also undermine control through opacity and variable output quality. Similarly, generative AI may increase or decrease perceived task value depending on whether learners view it as useful and aligned with learning goals or as distracting and superficial. Objectives and Purpose This meta-analysis pursues three objectives with direct relevance to European educational policy: First, to provide theoretically coherent synthesis enabling cross-national comparability essential for European educational cooperation and evidence-based decision-making across the EHEA. Second, to identify contextual moderators differentiating emotional outcomes—valuable for European institutions navigating tensions between continent-wide AI policies and nationally specific pedagogical traditions, enabling context-sensitive implementation respecting educational diversity. Third, to establish methodologically rigorous foundation informing European policy frameworks, teacher professional development, and institutional AI adoption strategies. By synthesizing global evidence, this work positions European educational research within worldwide developments while providing findings directly applicable to European contexts, strengthening Europe's leadership in responsible, evidence-based AI integration in education. Methodology, Methods, Research Instruments or Sources Used Search Strategy and Eligibility Criteria We conducted a systematic literature search following PRISMA 2020 guidelines across five international databases: Web of Science, Scopus, ERIC, IEEE Xplore, and PsychArchives. The search encompassed all studies published through December 2025, using a comprehensive search strategy organized into three concept groups: (1) generative AI (including ChatGPT, Gemini, Claude, large language models, AI chatbots); (2) education (including students, learners, K-12, higher education); and (3) emotions (based on Pekrun's taxonomy of achievement emotions). Within each group, terms were combined using Boolean OR, and groups were linked using AND. The search identified 5,182 records. After removing 2,605 duplicates, 2,577 unique records underwent title and abstract screening by two independent reviewers. Studies were included if they: (1) employed experimental or quasi-experimental designs with generative AI interventions and control conditions; (2) quantitatively measured at least one learner emotion; (3) were conducted in formal educational settings; (4) reported sufficient data for effect size calculation; and (5) were published in English in peer-reviewed venues. Data Extraction and Reliability Data extraction was independently performed by two researchers using a structured codebook. For each study, we extracted study identifiers, design characteristics, moderator variables (intervention time, educational level, academic discipline), emotion measures, and quantitative data (group means, standard deviations, sample sizes). Inter-rater reliability was rigorously assessed. For categorical variables (study design, educational level, discipline, intervention time, emotion type), Cohen's kappa coefficients ranged from 0.86 to 0.94, indicating excellent agreement. For continuous variables (means, standard deviations, sample sizes), intraclass correlation coefficients exceeded 0.95, indicating near-perfect consistency. All discrepancies were resolved through discussion until consensus. Statistical Analysis Effect sizes were computed as Hedges' g with small-sample bias correction. Given that multiple effect sizes were extracted from some studies (51 dependent effect sizes from 25 studies), we employed three-level random-effects meta-analytic models to account for statistical dependence. Level 1 represented sampling variance, Level 2 captured within-study variance among multiple effect sizes, and Level 3 represented between-study variance. Variance components were estimated using restricted maximum likelihood. Separate three-level models were fitted for each emotion category. Moderator analyses used three-level mixed-effects models to examine whether effects differed across intervention time, educational level, and academic discipline. Publication bias was assessed through funnel plot inspection and Egger-type regression with cluster-robust standard errors. Methodological quality was appraised using the Quality Assessment with Diverse Studies tool (ICC = 0.93). Conclusions, Expected Outcomes or Findings Main Findings The meta-analysis revealed differentiated emotional effects of generative AI use across all four theoretically defined emotion categories. Generative AI use was associated with significant increases in positive-activating emotions (g = 0.72, 95% CI [0.25, 1.19], p < .01) and positive-deactivating emotions (g = 0.57, 95% CI [0.24, 0.92], p < .001), indicating that learners experienced greater enjoyment and satisfaction in AI-supported learning conditions. Simultaneously, negative-activating emotions were significantly reduced (g = -0.47, 95% CI [-0.83, -0.12], p < .01), suggesting that generative AI use alleviated anxiety. The effect on negative-deactivating emotions was non-significant overall (g = -0.35, 95% CI [-0.90, 0.19]), but substantial heterogeneity (I² = 84.4%) pointed to context-dependent effects. Moderator analyses revealed that educational level and academic discipline did not significantly moderate emotional outcomes, suggesting that the emotional effects of generative AI are broadly consistent across K-12 and higher education settings and across STEM and non-STEM disciplines. However, intervention time significantly moderated effects on negative-deactivating emotions (QM = 5.23, p = .022), with longer interventions (≥10 weeks) associated with significant reductions in boredom, whereas shorter interventions showed no reliable effect. Theoretical and Practical Implications These findings provide robust empirical support for Control-Value Theory's applicability to generative AI-supported learning and demonstrate that emotional responses are shaped by learners' control and value appraisals rather than being inherent properties of the technology itself. Practically, the results indicate that generative AI can serve as an emotionally supportive instructional resource when implemented to enhance learners' perceived control and task value, while highlighting the importance of sustained, pedagogically integrated use for addressing deactivating negative emotions. The international scope of the evidence base ensures that these insights are relevant across diverse educational contexts worldwide. References Theoretical Foundation Pekrun, R. (2006). The control-value theory of achievement emotions: Assumptions, corollaries, and implications for educational research and practice. Educational Psychology Review, 18(4), 315-341. Pekrun, R. (2024). Control-Value Theory: From achievement emotion to a general theory of human emotions. Educational Psychology Review, 36(3), 83. Pekrun, R., Marsh, H. W., Elliot, A. J., Stockinger, K., Perry, R. P., Vogl, E., Goetz, T., Van Tilburg, W. A. P., Lüdtke, O., & Vispoel, W. P. (2023). A three-dimensional taxonomy of achievement emotions. Journal of Personality and Social Psychology, 124(1), 145-178. Methodological Foundation Page, M. J., McKenzie, J. E., Bossuyt, P. M., et al. (2021). The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. BMJ, 372, n71. Van Den Noortgate, W., López-López, J. A., Marín-Martínez, F., & Sánchez-Meca, J. (2013). Three-level meta-analysis of dependent effect sizes. Behavior Research Methods, 45(2), 576-594. Contemporary Context Ng, D. T. K., Chan, E. K. C., & Lo, C. K. (2025). Opportunities, challenges and school strategies for integrating generative AI in education. Computers and Education: Artificial Intelligence, 8, 100373. Wu, R., & Yu, Z. (2024). Do AI chatbots improve students learning outcomes? Evidence from a meta-analysis. British Journal of Educational Technology, 55(1), 10-33. Representative International Primary Studies Ismail, S. M. (2024). EFL learners' positive emotions in the era of technology: Unpacking the effects of artificial intelligence on learning enjoyment, self-efficacy, and resilience. Computer Assisted Language Learning Electronic Journal, 25(4), 526-551. Unal, Z. (2024). Transforming faculty-selected course materials into effective study tools: The role of AI in enhancing student learning and satisfaction. International Journal on E-Learning, 24(3), 223-240. Wang, Y. (2025). Reducing anxiety, promoting enjoyment and enhancing overall English proficiency: The impact of AI-assisted language learning in Chinese EFL contexts. British Educational Research Journal, berj.4187. Wu, Q., & Xu, A. (2025). Poe or Gemini for fostering writing skills in Japanese upper-intermediate learners: Uncovering the consequences on positive emotions, boredom to write, academic self-efficacy and writing development. British Educational Research Journal, berj.4119. Zhang, C., Meng, Y., & Ma, X. (2024). Artificial intelligence in EFL speaking: Impact on enjoyment, anxiety, and willingness to communicate. System, 121, 103259. 16. ICT in Education and Training
Paper The Inverted Digital Divide: How Generative AI Reconfigures Educational Inequality KU Leuven, Belgium Presenting Author:Research on the digital divide typically predicts that students from socio-economically advantaged groups adopt new educational technologies earlier, use them more “productively,” and translate these advantages into higher achievement. This proposal introduces and empirically examines an alternative possibility for the era of generative artificial intelligence (GenAI): an inverted digital divide, in which learners historically marginalized by parental education, migration background, or language status show higher educational use and larger academic gains from GenAI than their more privileged peers. Our core objective is to test whether GenAI’s distinctive affordances (especially multilingual interaction, low-friction conversational interfaces, and affirming dialogic feedback) alter the traditional sequence of digital inequality across three levels: (1) access, (2) educational use, and (3) academic outcomes. We ask:
To address these questions, we combine two complementary studies in the same national context. Study 1 maps adoption and educational uses of GenAI in a large sample of first-year university students across a wide range of programs, and links usage to early academic success. Study 2 tests causality via a randomized controlled trial in secondary education, estimating the effect of providing access to a high-capability GenAI tool during an authentic writing task and assessing whether treatment effects differ by students’ social background. By integrating large-scale descriptive evidence with experimental causal inference, this proposal aims to advance digital divide theory beyond “technology inevitably widens gaps,” specifying conditions under which certain designs and interactional norms may instead help underserved learners convert access into meaningful academic support. The Inverted Digital Divide: How Generative AI Reconfigures Educational Inequality Methodology, Methods, Research Instruments or Sources Used Design: Two-study mixed-methods design combining (a) a large-scale survey with multilevel modeling and (b) a cluster-randomized controlled trial. Study 1: University survey (access, use, outcomes). A survey was administered to 2,965 first-year students in 47 bachelor’s programs across 13 universities in Belgium (Flanders). Students reported GenAI use frequency (5-point scale from never to multiple times daily) and indicated whether they used GenAI for 11 academic activities (binary indicators such as explaining content, summarizing, proofreading, brainstorming, practice questions, exam preparation, literature search, and translation). Group differences by parental education, migration background, language background, and gender were tested using Welch’s t-tests with effect sizes. Academic outcomes were operationalized as the proportion of first-semester courses passed. To test whether GenAI use moderates achievement gaps, multilevel linear models (students nested in programs; random intercepts) estimated interaction terms between GenAI frequency and social-background indicators, controlling for gender, secondary-school track, and program type. Missing data were handled via multiple imputation. Study 2: Secondary-school RCT (causal effects and moderation). A cluster-randomized trial involved 280 students (ages 13–17) in four secondary schools. Classes were randomly assigned to a GenAI-access condition (tool available during writing) or a control condition (internet access without AI tools). Students completed a curriculum-relevant essay task. Outcomes were task performance (0–4; number of accurate, school-specific prevention examples) and language performance (0–4; grammar/spelling/coherence). Essays were scored by two human raters and two LLM assessors using a shared rubric; averaged scores were analyzed with OLS regression, including interactions between condition and parental education and migration background, with covariates (gender, prior achievement) and multiple imputation for missing data. Conclusions, Expected Outcomes or Findings Across two complementary studies, we find consistent evidence that GenAI may reshape (rather than reproduce) classic digital-divide patterns. First, GenAI access is near-saturated in the university sample, suggesting that in a high-connectivity setting, first-level divides in basic adoption can be small. Second, clearer differences emerge in educational uses: students from historically underserved groups (first-generation, migration-background, and language-minority learners) report more frequent and more learning-oriented GenAI use, particularly for language-intensive tasks (e.g., proofreading and translation) and studying support (e.g., practice questions). Third, the findings point toward an outcome-level pattern compatible with an “inverted digital divide.” In the survey, heavier GenAI use is associated with smaller achievement gaps, especially by parental education. In the randomized trial, providing GenAI access improves writing outcomes overall, and the gains are largest for specific underserved groups—notably stronger content gains among students with lower parental education, and stronger language gains among students with a migration background. Together, these results suggest that GenAI can function as a low-cost, on-demand scaffold resembling tutoring and language support, with disproportionate value for learners who have less access to such resources offline. Implications are conditional: GenAI’s equalizing potential depends on equitable access (including avoiding paywalled stratification), instruction that builds critical AI and prompt literacies, and assessment designs that reward synthesis and evaluation rather than verbatim generation. Limitations include reliance on self-reported use in Study 1 and task-specific generalizability in Study 2, motivating longitudinal and trace-data follow-ups. References Hargittai, E. (2002). Second-level digital divide: Differences in people’s online skills. First Monday, 7(4). van Dijk, J. (2020). The digital divide (2nd ed.). Polity. Van Deursen, A. J., & Helsper, E. J. (2015). The third-level digital divide: Who benefits most from being online? In Communication and Information Technologies Annual (Vol. 10, pp. 29–52). Emerald. Lythreatis, S., Singh, S. K., & El-Kassar, A. N. (2022). The digital divide: A review and future research agenda. Technological Forecasting and Social Change, 175, 121359. UNESCO. (2023). Guidance for Generative AI in Education and Research. UNESCO. Deng, R., Jiang, M., Yu, X., Lu, Y., & Liu, S. (2024). Does ChatGPT enhance student learning? A systematic review and meta-analysis. Computers & Education, 105224. Stephens, N. M., et al. (2012). Unseen disadvantage: University norms of independence and first-generation performance. Journal of Personality and Social Psychology, 102(6), 1178–1197. Walton, G. M., & Cohen, G. L. (2011). A brief social-belonging intervention improves outcomes of minority students. Science, 331(6023), 1447–1451. Sidoti, O., Park, E., & Gottfried, J. (2025). Teens’ use of ChatGPT for schoolwork. Pew Research Center. 16. ICT in Education and Training
Paper Perceived Digital Workload and Teacher Well-Being: The Mediating Role of Teacher Autonomy 1: Ankara University, Turkey (Türkiye); 2: Ministry of National Education (MoNE) Presenting Author:Problem Digitalisation in education systems has long been presented as a strategic transformation aimed at enhancing instructional processes, personalising learning, and rendering decision-making mechanisms more data-driven. The pedagogical potential of digital tools is commonly legitimised through discourses of accountability, monitoring, and quality assurance. Within this framework, educational organisations are increasingly governed through multiple digital platforms, databases, and reporting systems. However, the question of how this technically and administratively oriented digital transformation is reflected in teachers’ everyday work practices has often remained secondary. In particular, the growing expectations placed on teachers regarding digital data entry, reporting, and form completion alongside their pedagogical roles constitute a critical problem area concerning how digitalisation restructures teachers’ labour. Recent international literature emphasises that monitoring and reporting processes conducted through digital platforms not only increase teachers’ workload quantitatively but also transform the qualitative nature of teachers’ work. While digitalisation renders teacher performance more visible and measurable, it simultaneously generates new forms of workload that divert teachers’ time, attention, and emotional resources away from pedagogical interaction (Dormann ve diğerleri, 2019; Moraiti ve diğerleri, 2025). The systematic entry of student attendance data, the transfer of assessment results to digital platforms, the preparation of lesson plans, individual development reports, and the completion of numerous administrative forms in digital environments require teachers to allocate a substantial portion of their daily working time to non-pedagogical activities (Hase & Kuhl, 2024; Michos ve diğerleri, 2023). This situation entails the risk that instructional interaction, which constitutes the core of the teaching profession, is increasingly relegated to a secondary position. The literature commonly conceptualises the consequences of digitalisation in education through the notions of “digital overload” and “work intensification.” Digitalisation increasingly requires teachers to generate large volumes of data, enter this information into digital systems in a timely and accurate manner, and remain continuously visible through performance indicators. These demands contribute to heightened cognitive fatigue, intensified time pressure, and growing challenges in maintaining work–life balance, causing digital tools to function less as pedagogical supports and more as expanding domains of bureaucratic burden (Lepshokova ve diğerleri, 2025). This situation becomes particularly pronounced in education systems where multiple digital platforms are used simultaneously, as data duplication, technical disruptions, and unclear responsibility structures further intensify teachers’ perceived workload. Although digital competence, in-service training, and individual adaptation skills are frequently presented as mitigating factors, existing research indicates that such individual capacities are insufficient to compensate for structurally intensified work demands (Zhao ve diğerleri, 2025). Consequently, the core issue lies not primarily in teachers’ ability to use digital tools, but in how these technologies reorganise teachers’ labour and redefine professional roles within contemporary education systems. European Union education policies offer a comprehensive perspective that addresses this issue within the framework of teacher well-being and sustainable working conditions. The European Commission’s Digital Education Action Plan and teacher policy documents emphasise that digital transformation should be implemented in ways that strengthen teacher autonomy, support pedagogical capacity, and ensure a balanced management of workload (European Commission, 2020; European Commission, 2023). EU policy documents explicitly underline that excessive reporting and data demands pose significant risks to teacher motivation, professional satisfaction, and the long-term effectiveness of education systems (OECD, 2023). Accordingly, critically examining the effects of digitalisation on teachers’ workload remains a fundamental problem area not only in national contexts but also at the European level, particularly with regard to developing human-centred and sustainable education policies. The purpose of the present study is to examine the direct and indirect effects of teachers’ perceived digital workload on teacher autonomy and teacher well-being. Methodology, Methods, Research Instruments or Sources Used Method Research Design This study employed a correlational survey design to examine the relationships between teachers’ perceived digital workload, teacher autonomy, and teacher well-being. Within a theoretically grounded structural model, both the direct effects of perceived digital workload on teacher autonomy and teacher well-being and the mediating role of teacher autonomy in these relationships were tested. The study involved no instructional or organisational intervention and relied on teachers’ self-reported perceptions of digital work demands emerging from their everyday professional practices. Sample The sample consisted of primary, lower secondary, and upper secondary school teachers working in public schools in Türkiye. Due to institutional access constraints commonly encountered in large-scale educational research (OECD, 2023), a convenience sampling strategy based on voluntary participation was adopted (Creswell, 2014). Data were collected via an online questionnaire, ensuring participant anonymity and voluntariness. Ethical approval was obtained from the Ankara University Human Research Ethics Committee, and informed consent was secured from all participants prior to data collection. Data Collection Instruments Teachers’ perceived digital workload was measured using the Teachers’ Perceived Digital Workload Scale (TDWPS), developed by the researchers based on the literature on digital overload, digital bureaucratic demands, and work intensification. The scale captures teachers’ perceptions of data entry, reporting requirements, platform-based administrative obligations, as well as the volume and complexity of digitally mediated tasks. Teacher autonomy was assessed using the Teacher Autonomy Scale developed by Çolak and Altınkurt (2017), which measures perceived control over instructional decision-making, classroom practices, and professional judgement. Teacher well-being was measured using the Teacher Well-Being Scale developed by Collie (2015). All instruments were administered using a five-point Likert-type scale ranging from strongly disagree to strongly agree. Prior to testing the structural model, confirmatory factor analyses (CFA) were conducted to assess the validity of the measurement models. Data Analysis Data were analysed using structural equation modelling (SEM). Prior to model estimation, the data were screened for missing values, outliers, and violations of normality assumptions. Given the likelihood of multivariate non-normality in self-report data, the Robust Maximum Likelihood (MLR) estimation method was employed (Satorra & Bentler, 2010). Model fit was evaluated using multiple indices, including RMSEA, SRMR, GFI, CFI, and IFI, based on recommended cut-off criteria (Byrne, 1998; Hu & Bentler, 1999). The mediating role of teacher autonomy was tested using a bootstrap procedure with bias-corrected confidence intervals (Preacher & Hayes, 2008). Conclusions, Expected Outcomes or Findings Expected Outcomes The present study is expected to provide empirical evidence on the relationships between teachers’ perceived digital workload, teacher autonomy, and teacher well-being. It is anticipated that as perceived digital workload increases, teachers’ perceptions of control over professional decision-making processes will decrease, which in turn will negatively affect teacher well-being. Furthermore, teacher autonomy is expected to function as a significant mediating variable in the relationship between perceived digital workload and teacher well-being. This finding would indicate that the effects of digital workload on teacher well-being emerge partly through teachers’ perceived autonomy over their professional practices and decisions. Such a result would be consistent with theoretical approaches that conceptualise teacher autonomy as a core psychosocial mechanism explaining how digital working conditions are translated into well-being outcomes. From a broader perspective, the expected findings are likely to extend the existing literature, which has predominantly addressed the effects of digitalisation on teachers through the lenses of technical adaptation and individual competencies. By shifting the analytical focus toward the structural implications of digital transformation for teachers’ labour, working conditions, and well-being, the study aims to contribute to a more critical and comprehensive understanding of digitalisation in education. Modelling teacher autonomy as a central explanatory mechanism has the potential to generate evidence-based implications for the development of teacher-centred digital education practices and school-level policy processes that support teacher well-being. References References Byrne, B.M. (1998). Multivariate applications book series. Mahwah, NJ: Lawrence Erlbaum Associates Publishers. Creswell, J. W. (2014). Research design: Qualitative, quantitative, and mixed methods approaches (4th ed.). Sage. Dormann, M., Hinz, S., & Wittmann, E. (2019). Improving school administration through information technology? How digitalisation changes the bureaucratic features of public school administration. Educational Management Administration & Leadership, 47(2), 275–293. https://doi.org/10.1177/1741143217739362 European Commission. (2020). Digital education action plan (2021–2027): Resetting education and training for the digital age. Publications Office of the European Union. European Commission. (2023). Education and training monitor 2023. Publications Office of the European Union. Hase, A., & Kuhl, P. (2024). Teachers’ use of data from digital learning platforms for instructional design: A systematic review. Educational Technology Research and Development, 72, 1–23. https://doi.org/10.1007/s11423-023-10256-9 Hu, L., & Bentler, P.M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6(1), 1-55. Lepshokova, Z. K., Iashina, A. G., & Chernaya, A. V. (2025). Factors of digital overload among teachers. Population and Economics, 9(1), 45–63. Michos, K., Schmitz, M.-L., & Petko, D. (2023). Teachers’ data literacy for learning analytics: A central predictor for digital data use in upper secondary schools. Education and Information Technologies, 28, 15647–15666. https://doi.org/10.1007/s10639-023-11668-2 Moraiti, K., Bergviken Rensfeldt, A., & Lundin, M. (2025). Digital platform work reinforcing performativity: Teacher responses to work intensification explored through trace ethnography. Critical Studies in Education, 66(1), 1–17. OECD. (2023). Teachers and digital transformation: Working conditions, workload and well-being. OECD. https://doi.org/10.1787/teacher-digital-workload-2023 Preacher, K.J., & Hayes, A.F. (2008). Asymptotic and resampling strategies for assessing and comparing indirect effects in multiple mediator models. Behavior Research Methods, 40(3), 879–891. https://doi.org/10.3758/BRM.40.3.879 Satorra, A., & Bentler, P.M. (2010). Ensuring positiveness of the scaled difference chi-square test statistic. Psychometrika, 75(2), 243-248. https://doi.org/10.1007/s11336-009-9135-y Zhao, Y., Zhao, K., & Wei, S. (2025). School support, perceived value and teachers’ digital training adaptability: A multilevel moderated mediation model. Psychology in the Schools, 62(1), 45–62. https://doi.org/10.1002/pits.23001 16. ICT in Education and Training
Paper Learning Patterns of Computer and Information Literacy in Unexpectedly High-Performing Schools: An Analysis Based on ICILS 2023 Data from Europe Paderborn University, Germany Presenting Author:Computer and information literacy (CIL) has become a central educational objective across European education systems (European Commission, 2020), particularly considering fast-paced technological change and increasing societal reliance on digital technology. Accordingly, schools are mandated to systematically foster this key competence among all students (Fraillon, 2025). However, findings from the International Computer and Information Literacy Study 2023 (ICILS 2023) reveal persistent socio-economic disparities in CIL across Europe (European Commission, 2024; Fraillon, 2025; Kennedy et al., 2025). At the same time, research has identified so-called unexpectedly high-performing schools that have proven resilient despite serving student populations with low socioeconomic backgrounds and whose students demonstrate above-average performance in CIL (Drossel et al., 2020). Theoretical accounts of unexpectedly high-performing schools suggest that their high CIL levels can be attributed to an enhanced capacity to adapt to contextual conditions as well as to the diverse learning needs of their students (Muijs et al., 2004). Against this backdrop, learning-related factors may play a crucial role in buffering the adverse effects of socio-economic disadvantage in such schools (Niemann et al., 2025) and to learn CIL. Theoretically, CIL can be conceptualized as comprising four interrelated subdomains: (1) knowledge related to the use of computers, (2) the collection and organization of information, (3) the creation of digital content, and (4) digital communication (Duckworth & Fraillon, 2025). Although the CIL framework is well established, empirical research on the extent of learning across the four CIL subdomains including comparisons between unexpectedly high-performing schools and other schools has largely remained descriptive, typically examining the learning of all CIL-subdomains concurrently by assessing students’ extent of learning various internet-related tasks within school contexts by a scale. As a result of these overarching analyses across all CIL-subdomains, comparative examinations of selected European countries reveal no differences between unexpectedly high-performing schools and other schools (Niemann et al., 2025). This finding is surprising insofar as the theoretical assumptions mentioned suggest that learning, particularly in unexpectedly high-performing schools, should differ from those observed in other schools (Muijs et al., 2004; Niemann et al., 2025). Against this background, the present study aims to move beyond overarching approaches by adopting a person-centered perspective on student learning patterns taking all four CIL subdomains into account. The central objective is to identify unexpectedly high-performing schools and possible learning patterns of CIL among students attending those schools in comparison to other schools in selected European countries. Focusing on Germany, Finland, and the Czech Republic allows for a comparative analysis across three European education systems with differing CIL levels, instructional practices, school autonomy and taking highly different contextual conditions into account (Fraillon, 2025). Following on from this research gap, this presentation answers two research questions (RQ):
By addressing these questions, the study seeks to contribute to a more differentiated understanding of how learning CIL is taking place within unexpectedly high-performing schools across Europe. Methodology, Methods, Research Instruments or Sources Used Addressing the outlined research questions, empirical secondary analysis of data from the International Association for the Evaluation of Educational Achievement (IEA) International Computer and Information Literacy Study 2023 (ICILS 2023; Fraillon, 2025) were conducted in this contribution. The analyses are based on nationally representative drawn samples including eight-grade students from Germany (N = 5065), Finland (N = 4249) and the Czech Republic (N = 8169). Based on this sample, a two-step process was used to answer the research questions. In the first step (RQ1), unexpectedly high-performing schools were identified by aggregating the range of socio-economic status and the range of CIL of the students at the school level for each country. Unexpectedly high-performing schools are defined as those schools whose student body is in the bottom 40 percent of the socioeconomic status spectrum (as determined by the Highest International Socio-Economic Index of Occupational Status (HISEI)) and at the same time in the top 40 percent of the CIL range for the respective country (Drossel et al., 2024). All other schools within the sample are regarded as other schools. In a second step (RQ2), latent class analyses (LCA) using MPlus 8 (Muthén & Muthén, 2017) were conducted utilizing the respective subsamples (students at unexpectedly high-performing schools; Germany: n = 520; Finland: n = 505; Czech Republic: n = 411 vs. students from other schools; Germany: n = 4503; Finland: n = 3728; Czech Republic: n = 7758) from the three countries. Missing values were excluded from the analyses. To identify students distinct learning patterns, the various CIL areas were operationalized with a total of 10 items relating to what extent students have learned how to do different internet- and ICT-related tasks at school (for example Use the internet to find information, Judge whether a message from someone is a scam, Edit the layout and formatting of documents or slideshow presentations). The LCA allows the identification of unobserved subgroups of students with similar patterns across the extent of learning of the four CIL subdomains. Model selection follows established statistical criteria, including the Bayesian Information Criterion (BIC) (Eshima, 2022; Nylund-Gibson & Choi, 2018). Sampling weights were applied to account for potential sampling bias (Tieck, 2025) and the complex survey design of ICILS 2023 are considered to ensure the representativeness and robustness of the results (Fraillon, 2025). Conclusions, Expected Outcomes or Findings The analyses for RQ1 revealed that in Finland 13 percent of students attend unexpectedly high-performing schools, compared with 7 percent in the Czech Republic and 14 percent in Germany. Moreover, to answer RQ2 three distinct and consistent patterns of students’ CIL learning across all countries and both unexpectedly high-performing and other schools were identified. Owing to strong similarities, the patterns were labelled consistently. All-around CIL learners demonstrated high engagement across all CIL subdomains. They constituted a substantially larger share of students in unexpectedly high-performing schools (Czech Republic: 50%; Finland: 43%; Germany: 58%) than in other schools (34%, 33%, and 40%, respectively), while showing comparable engagement profiles across school types. Selective CIL learners focused on a narrower but still substantial range of CIL-related activities, particularly online information search and document or presentation formatting. In unexpectedly high-performing schools, their proportions were lower (Czech Republic: 29%; Finland: 33%; Germany: 10%) than in other schools, where Selective learners represented a larger share of students (47%, 52%, and 23%, respectively), despite displaying similar learning patterns. Reserved CIL learners showed generally low engagement across CIL subdomains, concentrating mainly on basic activities. They accounted for around one quarter to one third of students in unexpectedly high-performing schools (Czech Republic: 28%; Finland: 17%; Germany: 33%) and for comparable shares in other schools (20%, 15%, and 37%, respectively). Overall, the findings indicate the existence of unexpectedly high-performing schools in all three countries with a stable typology of CIL learning patterns. From a research perspective, the results highlight the need for systematic investigation of school-level conditions influencing the emergence and distribution of CIL learning patterns. From a school perspective, the higher concentration of All-around CIL learners in unexpectedly high-performing schools underscores the role of school practices and learning environments in fostering comprehensive CIL engagement, with implications potentially transferable across countries. References Drossel, K., Eickelmann, B. & Vennemann, M. (2020). Schools overcoming the digital divide: in depth analyses towards organizational resilience in the computer and information literacy domain. Large-scale Assess Educ, 8(9). https://doi.org/10.1186/s40536-020-00087-w Drossel, K., Oldak, A., Bette, R. & Eickelmann, B. (2024). Unexpectedly successful! Results of the UneS project and recommendations for action for equal opportunities and digitalisation-related school development processes at individual school level. [Unerwartbar erfolgreich! Ergebnisse des UneS-Projektes und Handlungsempfehlungen für chancengerechte, digitalisierungsbezogene Schulentwicklungsprozesse auf Einzelschulebene]. Waxmann. Duckworth, D. & Fraillon, J. (2025). Measuring Students’ Computer and Information Literacy. In J. Fraillon (Ed.), An International Perspective on Digital Literacy. Results from ICILS 2023 (pp. 71–87). Springer. Eshima, N. (2022). An Introduction to Latent Class Analysis (Vol. 14). Springer. European Commission (2020). Digital Education Action Plan 2021–2027 – Resetting education and training for the digital age. EUR-Lex. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A52020DC0624 European Commission (2024). International Computer and Information Literacy Study (ICILS) in Europe – 2023. Main Findings and Educational Policy Implications. Publications Office of the European Union. https://data.europa.eu/doi/10.2766/5221263 Fraillon, J. (Ed.). (2025). An International Perspective on Digital Literacy. Results from ICILS 2023. Springer. Kennedy, A., Strello, A., & Strietholt, R. (2025). Relationships between computer and information literacy, computational thinking and student background. In J. Fraillon (Ed.), An international perspective on digital literacy: Results from ICILS 2023 (pp. 155–177). IEA. Muijs, D., Harris, A., Chapman, C., Stoll, L. & Russ, J. (2004). Improving Schools in Socioeconomically Disadvantaged Areas – A Review of Research Evidence. School Effectiveness and School Improvement, 15(2), 149-175. http://dx.doi.org/10.1076/sesi.15.2.149.30433 Muthén, L.K. & Muthén, B.O. (2017). Mplus User’s Guide. Eighth Edition. Muthén & Muthén. Niemann, J., Eickelmann, B. & Drossel, K. (2025). Overcoming Digital Inequalities – Identification and Characterisation of Digitally Resilient Schools in Different Countries Using ICILS 2023 Data. Education Science, 15(7), 898. Nylund-Gibson, K., & Choi, A. Y. (2018). Ten frequently asked questions about latent class analysis. Translational Issues in Psychological Science, 4(4), 440–461. https://doi.org/10.1037/tps0000176 Tieck, S. (2025). Sampling weights and participation rates. In J. Fraillon, M. Rožman, S. Meyer, L. Musu, Y.-L. Liaw, A. Christiansen & S. Tieck (Hrsg.), International Computer and Information Literacy Study 2023. Technical Report (S. 89 – 112). IEA. | ||