Abstract
The past decade has witnessed a marked shift in English Language Teaching (ELT) assessment within higher education from standardised testing toward formative and learner-centred approaches. This study maps and reinterprets this evolution through a bibliometric and AI-assisted thematic analysis of 516 peer-reviewed journal articles published between 2014 and 2024. Using Dimensions.ai as the data source, co-word mapping and Latent Dirichlet Allocation (LDA) modelling identified five thematic clusters representing teacher assessment literacy, learner-centred experimentation, systemic reform, inclusivity, and sociocultural feedback dynamics. Although publication output increased exponentially after 2019, conceptual integration between sociocultural learning theory and validation frameworks remains limited. To address this gap, the study introduces relational validity as a synthesising framework that aligns evidential reasoning with dialogic, ethical, and pedagogical dimensions of assessment. This construct reframes validity as coherence among evidence, interpretation, and educational consequence, extending traditional psychometric concerns toward a more contextually grounded, human-centred paradigm. By combining structural bibliometric analysis with interpretive theorisation, this study offers both an overview of alternative assessment research and a conceptual roadmap for advancing fairness, reliability, and transformative practice in ELT higher education.