Large-scale semantic mapping of learner agency and autonomy reveals what measurement and generative AI research overlook

Researchers conducted a large-scale semantic analysis of 8,954 definitions and 2,700 scale items across 14,000+ publications to map how learner agency and autonomy are conceptualized and measured. They identified three core dimensions (task regulation, intrinsic motivation, and sociocultural action) and found that existing measurement scales systematically underrepresent the sociocultural aspect, while current generative AI applications in education narrowly focus on learning control.

This research addresses a fundamental problem in educational science: the inconsistent definition and measurement of learner agency and autonomy across studies. By analyzing over 14,000 publications through computational semantic mapping, the authors quantified the "jingle-jangle fallacy"—where researchers use identical terms to mean different things or different terms to mean the same thing. This fragmentation has prevented cumulative knowledge building and obscured what these constructs actually entail.

The study reveals that learner agency and autonomy operate across three distinct dimensions: task-level regulation and control, person-centered intrinsic motivation and decision-making, and sociocultural relational action. Current measurement scales disproportionately emphasize the first two dimensions while neglecting the sociocultural aspect, suggesting researchers have been measuring an incomplete picture of these constructs. This gap has real consequences for education technology development.

Generative AI applications in educational settings currently concentrate on learning regulation and control—precisely the dimensions that dominate existing scales. This narrow focus means AI-mediated learning environments are designed to optimize only part of what learner agency encompasses, potentially missing opportunities to cultivate social interaction, peer collaboration, and contextual autonomy. The research suggests that educational technologists and AI developers are inadvertently perpetuating measurement bias by building systems aligned with existing, incomplete conceptualizations rather than the full multidimensional nature of agency and autonomy.

This work carries implications beyond academia: education platforms, learning management systems, and AI tutors built on this incomplete understanding may fail to develop learners' full capacity for self-directed, socially-embedded learning.

• →Semantic analysis of 14,000+ publications reveals inconsistent definitions of learner agency and autonomy, creating a cumulative knowledge problem in education research.
• →Three distinct dimensions structure these constructs: task regulation, intrinsic motivation, and sociocultural action, with existing scales underrepresenting the social dimension.
• →Current generative AI in education concentrates on learning control and regulation, narrowing the behavioral capabilities designed into AI-mediated learning environments.
• →Measurement bias in existing scales directly influences how education technology and AI systems are conceptualized and developed.
• →Addressing this definitional and measurement gap could expand AI applications to better support multidimensional learner autonomy and social-relational learning.