Paper by Theo Berger & Hannes Scheffter: “This study provides an innovative approach to identify the relevant competence profiles that data science education should promote for work in the social domain. We assess the supply side of European Data Science for Good (DSG) initiatives and analyse 335 projects. Based on project documentation, we identify 146 distinct data science methods and develop a taxonomy that distinguishes between statistical and machine learning approaches and classifies methods according to the scale of the target variables (nominal, ordinal, metric).
The results show a strong dominance of machine learning, particularly in classification-oriented problem settings, alongside a substantial role for exploratory and descriptive analysis in evaluation tasks. Drawing on these patterns, we provide a discussion on the implications for curriculum design of higher education programmes in social data science and artificial intelligence, arguing for an educational profile that aligns advanced methodological skills in machine learning with solid statistical literacy…(More)”.
