Explainable Decision Intelligence for Lecturer Competency in AI-Enabled Higher Education

Universities are asking lecturers to work with AI-supported teaching tools, learning management systems, research-information platforms, and digital administrative dashboards. Yet lecturer readiness in such environments cannot be reduced to technical skill. This study evaluates an explainable Decision Intelligence framework for multiclass prediction of lecturer competency using DataSet.csv, a semicolon-delimited constructed/simulated analytical dataset of 500 anonymized lecturer profiles. The file is not an external higher-education registry and is not treated as independently verified institutional evidence. The framework compares a digital-only feature set with a socio-technical configuration containing six digital indicators, six psychological-readiness variables, and two workload measures. A stratified 80:20 train-test split and stratified 10-fold cross-validation were used. Adding psychological readiness and workload variables improved Logistic Regression accuracy from 0.77 to 0.87, Macro-F1 from 0.593 to 0.760, and macro ROC-AUC from 0.855 to 0.958; McNemar's exact test showed a significant paired improvement (p = 0.01294). Hybrid Soft Voting obtained the highest held-out macro ROC-AUC (0.965), whereas socio-technical Logistic Regression gave the strongest held-out accuracy and Macro-F1. SHAP ranked Motivation, Adaptability, Change Readiness, Emotional Intelligence, and Resistance to Change as leading global contributors, but these attributions explain model behavior rather than causal effects. The dashboard maps prediction confidence and SHAP-identified barriers to lecturer-development options, including AI-literacy support, learning-management-system guidance, pedagogical mentoring, research mentorship, peer learning, workload discussion, and academic HR review. It is not designed for automated lecturer evaluation, ranking, discipline, or academic promotion decisions. The findings should be read as internal framework validation; external, longitudinal, and higher-education-specific validation with independently labelled lecturer data is required before implementation.