From Knowing a Formula to Transferring Knowledge: A Human-Centred Interdisciplinary Model with Artificial Intelligence-Assisted Verification for Technical and Vocational Education

Author preprint. This manuscript has not been peer reviewed. This study presents an author-designed interdisciplinary transfer model for technical and vocational education and training. The model treats structural recognition, rather than formula reproduction, as the mechanism of transfer and organises learning through invariant mathematical structures, changing disciplinary contexts, practice-oriented tasks, and a three-level Contextual Heterogeneous Test. Two independent first-year cohorts were studied in a quasi-experimental design: 106 students in 2024 and 93 in 2025. In 2024, experimental and control groups were comparable on entry mathematics and physics measures; the transfer-block mean was 9.40 versus 4.61, respectively. In 2025, the transfer advantage was reproduced, 8.58 versus 5.46, but the experimental group entered with significantly stronger mathematics and physics results; therefore, the 2025 cohort is interpreted as replication of the direction of the transfer advantage rather than as an independent causal estimate. The assessment comprised 24 tasks in three blocks that progressively increased contextual distance. Cronbach alpha calculated across the three block scores was 0.60 in 2024 and 0.77 in 2025. Building on the empirical transfer architecture, the paper proposes a human-centred artificial intelligence extension in which generative AI is introduced only after an independent solution attempt and is used for comparison, error detection, verification, and justification. AI is therefore treated as a prospective verification layer, not as the cause of the reported 2024–2025 outcomes.