Predicting Future Skills through Competence Mining

Challenge

Universities are under growing pressure not only to keep their programmes academically current, but to strategically align them with the competencies of the future. The problem: module handbooks exist as unstructured, inconsistent text data – difficult to compare, barely machine-readable, and enormously labour-intensive to analyse manually. Strategic gap analyses – for instance, whether a curriculum adequately prepares students for Data & AI Literacy or Future Skills – therefore tend to remain qualitative, ad hoc, and without a reliable data foundation. Implicit competency assumptions embedded in learning objectives and module descriptions stay invisible, and with them, untapped potential for curriculum development.

Approach

STAT-UP developed a three-stage Competence Mining Pipeline that does more than describe curricular content – it models it as a structured competency network. In the first stage, module handbooks are transformed into a Knowledge Graph using NLP and

Large Language Models: modules, competencies, methods, and roles are extracted as nodes, normalised, and connected through semantic relations. In the second stage, Link Prediction identifies which competency connections are structurally implied but not yet explicitly visible in the curriculum – generating data-driven hypotheses about bridge competencies. In the third stage, the results are triangulated via Path Mining against external reference frameworks: the Stifterverband's Future Skills 2030 framework, the Data & AI Literacy model by Schüller et al., and the IEEE Draft Standard for Data and AI Competency. The result is a systematic map of existing strengths, implicit connections, and blind spots within the curriculum.

Impact

Applied to a model module handbook for the fictitious programme "Civic Transformation, Data, and AI for Public Services", the pipeline revealed clear patterns: Community-oriented and Digital Future Skills were already well anchored with a coverage ratio of 0.80 each, while Transformative Future Skills showed significant gaps at 0.33. Link Prediction identified approximately 28 new potential competency connections – including bridges from concrete course content such as "Inclusive Digital Service Design" to overarching future competencies like ethical reasoning and responsibility & accountability. With a recall of 0.714, the method surfaces the majority of relevant connection points, providing a structured, data-driven foundation for strategic curriculum discussions – without replacing the expert judgement of faculty.