Background

The rapid diffusion of generative artificial intelligence in higher education has raised significant pedagogical, ethical, and epistemic concerns across Europe (OECD, 2026). European policy initiatives, including the European Digital Education Action Plan and ongoing debates surrounding the EU AI Act, explicitly emphasize the need to strengthen AI literacy, critical thinking, and responsible AI use in educational contexts (Digital Education Council, 2024). While these frameworks highlight the societal relevance of critical engagement with AI, higher education institutions often lack empirically grounded didactic models that translate such policy goals into concrete teaching practices (Han et al., 2025). As a result, there remains a gap between European-level policy aspirations and evidence-based instructional approaches that support students in critically engaging with GenAI tools in academic learning.

Research gap

Research consistently identifies critical thinking as a central competence in higher education and demonstrates that it can be fostered through targeted instructional interventions (Abrami et al., 2008; Halpern, 1998). At the same time, a growing body of literature addresses AI literacy and the educational implications of generative AI, discussing both its potential to support learning and its risks for superficial or uncritical engagement. However, much of this research remains conceptual, focuses predominantly on K–12 contexts, or relies on survey-based quantitative approaches.

Recent reviews (e.g., Chara-De lo Rios et al., 2025; Zhai et al., 2024) identify a clear lack of empirically evaluated didactic models that foster critical thinking in higher education AI contexts, particularly within European settings. Moreover, few studies investigate students’ real-time reasoning processes while interacting with GenAI tools. By adopting a qualitative, process-oriented perspective, this study directly addresses these gaps and contributes empirical evidence on how critical thinking unfolds during students- GenAI interaction. 

Theoretical Framework

The theoretical framework of this study is primarily based on the integrative model of critical thinking proposed by Dwyer et al. (2014), which conceptualises critical thinking as a set of interrelated cognitive processes involved in reasoning and judgement. The framework emphasises four key dimensions: analysis, evaluation, inference and reflective judgement, which are particularly relevant in complex and ill-structured problem contexts. Building on the Delphi Report (Facione, 1990), these dimensions involve identifying argumentative structures, evaluating the credibility and coherence of information, drawing reasoned conclusions and reflecting on uncertainty and the provisional nature of knowledge.

To operationalize critical thinking in AI-related learning environments, the study draws on the didactic framework developed by Faust and Mayweg-Paus (2024). This model structures critical reflection along the machine learning life cycle and uses guiding questions related to transparency, ethics, reliability, trustworthiness, security, control, and data protection. In the present study, this framework serves both as the instructional intervention and as an analytical lens for examining how students critically engage with GenAI systems.

Research Questions

This study examines how a didactic model for critically questioning machine learning systems across their life cycle influences university students’ critical thinking when interacting with generative AI (GenAI) tools. The main research question is: In what ways does instruction based on this model shape students’ critical engagement with GenAI outputs?

Two sub-questions guide the analysis. First, the study explores how students with and without prior instruction in the model differ in their cognitive critical thinking skills engagement with GenAI-generated content. Second, it investigates which additional forms of reflective behavior, such as prompting strategies or attitudes toward AI outputs, emerge in students’ think-aloud processes during interaction with GenAI tools.

The study employs a quasi-experimental mixed-methods design with two cohorts of undergraduate students at a German research university. One cohort completed AI-based reflective tasks without prior instruction in the didactic model (control group), while the second cohort received explicit model-based instruction during a seminar on critical thinking (experimental group). Data consist of written Think-Aloud Protocols, in which students documented their thoughts, prompts, evaluations, and decisions while using generative AI tools ((excluding ChatGPT) to examine argumentative theses. The material was analyzed using qualitative content analysis supported by MAXQDA, combining deductive coding based on critical thinking dimensions (analysis, evaluation, inference, self-regulation, reflective judgement) with inductive category development. Complementary quantitative frequency analyses were used to compare patterns across groups.

The findings indicate that students who received model-based instruction demonstrate more differentiated analytical reasoning, greater awareness of uncertainty, and stronger reflective judgement when engaging with GenAI outputs. They show increased skepticism, more deliberate prompt revision strategies, and more explicit distancing from AI-generated claims. Beyond predefined critical thinking skills, the analysis reveals emergent categories related to prompting behavior and attitudes toward AI, highlighting critical engagement as a dynamic, process-oriented competence. Overall, the results suggest that structured, life-cycle-based reflection supports students in developing critical and responsible AI use. The study contributes empirical evidence to European debates on AI literacy and critical thinking in higher education. It demonstrates how theoretically grounded didactic models can be operationalized and evaluated in real classroom settings. Implications are discussed for curriculum design, teacher education, and future cross-national research on critical thinking and AI in higher education.

Abrami, P. C., Bernard, R. M., Borokhovski, E., Wade, A., Surkes, M. A., Tamim, R., & Zhang, D. (2008). Instructional Interventions Affecting Critical Thinking Skills and Dispositions: A Stage 1 Meta-Analysis. Review of Educational Research, 78(4), 1102-1134. https://doi.org/10.3102/0034654308326084 Chara-De lo Rios, T., Solis-Trujillo, B., Perez-Ruiz, J., & Aquije-Mansilla, M. (2025). Systematic review of critical thinking using artificial intelligence. Edelweiss Applied Science and Technology, 9(3), 990–1001. Digital Education Council. (2024, 17. März). EU AI Act: What it means for universities [Executive briefing]. Digital Education Council. https://www.digitaleducationcouncil.com/post/eu-ai-act-what-it-means-for-universities Dwyer, C. P., Hogan, M. J., & Stewart, I. (2014). An integrated critical thinking framework for the 21st century. Thinking Skills and Creativity, 12, 43–52. https://doi.org/10.1016/j.tsc.2013.12.004 Facione, P. A. (1990). The California Critical Thinking Skills Test: College level. Experimental validation and content validity. California Academic Press. Faust, A., & Mayweg-Paus, E. (2024). Empowering Tomorrow’s Minds. In H. Crompton & D. Burke, Artificial Intelligence Applications in Higher Education (1. Aufl., S. 73–89). Routledge. https://doi.org/10.4324/9781003440178-5 Halpern, D. F. (1998). Teaching critical thinking for transfer across domains: Disposition, skills, structure training, and metacognitive monitoring. American Psychologist, 53(4), 449–455. https://doi.org/10.1037/0003-066X.53.4.449 Han, B., Nawaz, S., Buchanan, G., & McKay, D. (2025). Students’ perceptions: Exploring the interplay of ethical and pedagogical impacts for adopting AI in higher education. International Journal of Artificial Intelligence in Education, 35(1), 1–24. https://doi.org/10.1007/s40593-024-00456-4 Mayring, P. (2012). Qualitative Inhaltsanalyse–ein beispiel für mixed methods. Mixed Methods in der empirischen Bildungsforschung, 1, 27-36. Mayring, P. (2015). Qualitative Inhaltsanalyse: Grundlagen und Techniken (12. Aufl.). Beltz. OECD (2026), OECD Digital Education Outlook 2026: Exploring Effective Uses of Generative AI in Education, OECD Publishing, Paris, https://doi.org/10.1787/062a7394-en. Zhai, C., Wibowo, S., & Li, L. D. (2024). The effects of over-reliance on AI dialogue systems on students’ cognitive abilities: A systematic review. Smart Learning Environments, 11(1), 28. https://doi.org/10.1186/s40561-024-00316-7