The self-regulatory paradox of learning analytics: student expectations and the conditions for fair algorithmic assessment in higher education

Abstract

Learning Analytics (LA) are increasingly positioned as tools to foster self-regulated learning (SRL) in higher education. As assessment-related data are progressively processed, interpreted, and acted upon through algorithmic systems, the fairness, transparency, and perceived legitimacy of these systems depend not only on their technical design but on whether students recognise themselves as agents within them. Yet students’ own expectations regarding the institutional use of LA remain underexplored, particularly in face-to-face university contexts. This study investigates the service and feature expectations of 1,020 undergraduate students at a large Spanish research university, using the Student Expectations of Learning Analytics Questionnaire (SELAQ). Descriptive analyses and between-group comparisons were applied to seven items rated on a five-point Likert scale. Findings reveal a consistent asymmetry: students most strongly expect professors to act on analytics when academic risk is detected, and show comparatively lower expectations for LA to support their own decision-making. This pattern—expecting LA to function as a teacher-mediated intervention tool rather than a learner-directed resource—constitutes what we term a self-regulatory paradox: students endorse the outcomes of self-regulated learning while simultaneously delegating its data-informed enactment to their instructors. A direct within-student comparison confirmed that instructor- and institution-oriented expectations exceeded student-oriented ones (paired-samples test, medium effect). Gender-based differences were statistically significant for all seven items but small in size and, after Bonferroni correction, robust for five of them (Cohen's d = 0.18–0.29), female students consistently reporting higher expectations. These findings have direct implications for the design of LA dashboards, data literacy training, and institutional strategies aimed at repositioning students as active, data-informed agents of their own learning. More broadly, the paradox specifies an empirical precondition for fair and transparent algorithmic assessment: systems cannot be considered legitimate by virtue of being technically accessible to learners if those learners do not recognise themselves as entitled to interpret and act upon their own assessment data. We argue that perceived fairness and transparency in automated and AI-assisted assessment are contingent on building, rather than presupposing, students’ data-informed agency.