Rethinking methodological foundations in Science Education Research in the age of artificial intelligence and big data

Artificial Intelligence (AI) and data science are no longer peripheral instruments in Science Education Research (SER). They offer new ways to organise, analyse, and produce knowledge. Computational approaches to textual data, including natural language processing, machine learning, and network analysis, provide models for addressing complex textual corpora, opening new possibilities for integrating qualitative sensitivity with quantitative scope. However, adopting AI-based methods is not merely a technical innovation but entails a profound shift in methodological assumptions. Two main gaps emerge: these assumptions remain largely underexplored, and we still lack a clear understanding of how AI-based methods relate to existing paradigms in SER. Moreover, the field lacks systematic instruments for analysing such methodological shifts and positioning them within the broader paradigmatic landscape. To address this gap, this thesis develops a methodological and philosophical reflection on how computational and AI-based approaches interact with qualitative inquiry in SER. Through two case studies on text analysis, it examines how these methods operate in practice and what kinds of knowledge they produce. From these analyses, a heuristic tool was designed to make visible the assumptions that govern research methods, structured along ontological, epistemological, and axiological dimensions, and subsequently applied to AI-driven methods in contemporary scientific research and to emerging hybrid approaches in science education. The results show that AI does not simply accelerate existing ways of producing scientific knowledge but expands them, acting as an epistemic agent that places data curation and manipulation at the centre of research. Furthermore, the combination of computational and qualitative approaches in SER proves most fruitful when their complementary features are valued within an iterative dialogue. By making these transformations explicit, the thesis contributes to a rethinking of mixed methods as a paradigmatic rather than procedural concept, and to a renewed vision of methodological pluralism in the age of Artificial Intelligence.