Applications of eXplainable Artificial Intelligence in Public Employment Services Decision Support Systems

The study explores the integration of eXplainable Artificial Intelligence (XAI) systems into the Decision Support Systems (DSS) of Public Employment Services (PES) to enhance the allocation of resources for reducing unemployment. The research focuses on profiling and clustering unemployed individuals in Switzerland using both supervised and unsupervised Machine Learning (ML) models. Traditional models like Logistic Regression and Decision Trees have been commonly used in PES but are often limited by their simplicity and limited interpretability. This study introduces more advanced models, such as XGBoost and Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN), to improve accuracy and clustering efficiency. The integration of XAI into these models ensures that predictions and decisions are interpretable, allowing caseworkers to make informed decisions. The XAI integration consists of three main components: a preprocessing pipeline, a clustering module, and a predicting module, which outputs explanations by design. The findings indicate that the XA enhanced models can provide actionable insights at the individual, cluster, and global levels, improving the efficiency of resource allocation and reducing long-term unemployment. Despite some limitations, including data quality and model performance, the study contributes to the literature by bridging the gap between AI in PES and XAI in DSS, and by implementing innovative clustering techniques within PES.