Stabilizing truth in educational sciences: a systematic review of generative AI in education

Purpose This paper examines how the scientific system, particularly the educational sciences, observes and communicates the phenomenon of generative artificial intelligence (AI) in education. Drawing on Luhmann’s social systems theory, the purpose is to reconstruct how scientific communication stabilizes truth by framing AI through distinctions such as augmentation, automation, and hybrid models. Design/methodology/approach A systematic literature review was conducted following the PRISMA 2020 framework, using Scopus to cover the period 2020–2025. Out of 1,105 initial publications, 110 studies met the inclusion criteria. Quality was assessed with the Mixed Methods Appraisal Tool (MMAT). Findings Educational sciences primarily frame generative AI as augmentation (71.8%), with automation (18.2%) and hybrid models (3.6%) secondary. Higher education dominates as the site of observation (65.5%), while primary and vocational education remain marginal. Ethical concerns focus on privacy, bias, accountability, and teacher autonomy. The scientific system reduces uncertainty by coding AI as supportive rather than disruptive. Research limitations/implications The review relied on Scopus, English-only full-text sources, and citation-based selection, which may bias findings. Practical implications The study enriches systematic review methodology by integrating systems theory, showing how reviews can serve not only as syntheses but also as tools for second-order observation of scientific communication. Originality/value By combining systematic review with systems theory, the study demonstrates how science programs its communication about AI, offering a transferable model for analysing emerging technologies in other social systems.