• ats
• fairness
• interpretability
• knowldege bases
• learning to rank
• matching algorithms
• screening
• skill assessment
• xai

Applicant Tracking Systems (ATSs) speed up the hiring process, particularly during the screening phase, by analyzing resumes and ranking candidates. The ranking is determined by assigning an individual score to each candidate who applies for a given job position to identify the most suitable to be promoted in the selection process.
This thesis proposes a set of tools to integrate into an ATS. We propose a synthetic data generator to overcome privacy and data collection issues. Then, we offer an Interpretable-by-design Human Recommender (IHR) framework: starting from job requirements and curricula’s traits in tabular format, IHR applies a set of matching functions to derive a fitness matrix; then, a transparent model uses the matrix to predict the corresponding suitability scores. In addition, the IHR offers visual and textual multi-stakeholder explanations to elucidate why the candidate obtained the score. Finally, we designed a tool that helps candidates express their skills, avoiding matching errors. All tools rely on Knowledge Bases to improve data generation, matching accuracy, and explanation quality. The proposed tools emphasize transparency, fairness, and flexibility.
We conduct experiments to show the synthetic data generator benefit and highlight how the IHR brings out the bias. Supported by the obtained results, we suggest that interpretability increases the efficiency and fairness of the hiring process by enabling decision-makers to rely on transparent criteria.