This study investigates how universities can recognize and assess students’ actual learning processes in an era where generative artificial intelligence has become a primary medium of learning.
Focusing on cyber universities, it identifies a structural limitation in existing academic systems, which rely predominantly on exam scores and final products and therefore fail to capture learners’ cognitive trajectories and evolving understanding.
To address this gap, the study proposes TB-AIMS (Token-Based AI Academic Management System), a novel academic architecture that connects students’ personal AI learning environments with institutional academic systems.
TB-AIMS interprets course-specific AI workspaces, dialogue logs, and token usage records as formal learning evidence.
Grounded in Evidence-Centered Design and Stealth Assessment, this approach reconceptualizes learning not as isolated outcomes but as a continuous trajectory of thinking, questioning, and refinement.
TB-AIMS introduces a Learning Record Layer that stores AI-mediated learning traces independently of specific AI platforms, enabling privacy-preserving, vendor-neutral integration with university learning management and assessment systems.
A token-based resource model further reframes AI usage as a strategic learning activity rather than mere consumption, supporting process-based indicators such as conceptual evolution, token efficiency, and self-regulation.
By repositioning generative AI from a threat to academic integrity to an officially recognized learning partner, TB-AIMS redefines the role of universities from content providers to interpreters and certifiers of learning processes.
The framework offers a new paradigm for academic assessment and learner agency in converging online and offline higher-education environments shaped by AI.