The LIPS Platform: A Seven-Layer Ontological Model for Micro-Skill Assessment and Adaptive Recommendation Supported by Generative AI

The integration of artificial intelligence (AI) into health systems requires instruments capable of diagnosing student performance beyond global scores, explaining why a failure occurs and how to remediate it, without delegating decisions to opaque algorithms. This study, guided by Design Science Research, presents the design and specification of the LIPS platform (Integrated Laboratory of Simulated Practices), a closed-loop, seven-layer sociotechnical architecture for the granular assessment of competencies in health professions education. The architecture decomposes clinical activities into observable micro-skills, paired with assessment items through a Q-matrix, and isolates the deterministic computation of scores from interpretation by generative AI, restricted to the formulation of explanatory hypotheses linked to evidence and submitted to instructor validation (human-in-the-loop). The demonstration applied the method to an illustrative airway suctioning scenario: the decomposition generated ten micro-skills and twenty items fully linked through the Q-matrix, and the diagnostic pathway showed how a global performance of 90% would mask a critical safety failure localized in a specific micro-skill. The formative evaluation is presented as a structured protocol, prior to application with human participants. The central contribution lies in the method for granularity of competencies and for containment of generative AI, transferable to other health professions education systems.