Explainable person-job recommendations: challenges, approaches, and comparative analysis.

Area of Science:

• Artificial Intelligence
• Human-Computer Interaction
• Information Science

Background:

• Person-job recommendation systems (PJRS) are increasingly used in hiring, raising concerns about their "black box" nature.
• Demand for explainable AI (XAI) solutions is growing to address the opacity of PJRS.
• This review systematically examines XAI methods applied to PJRS.

Purpose of the Study:

• To review and categorize XAI techniques for PJRS.
• To develop an integrated framework for transparent PJRS.
• To provide comparative evidence on the performance-explainability trade-offs of different XAI methods.

Main Methods:

• Systematic review of 85 studies (2019-2025) from major academic databases, following PRISMA 2020 guidelines.
• Categorization of explainability techniques into data, model, and output layers.
• Quantitative benchmarking of six representative XAI methods.

Main Results:

• XAI techniques were categorized into data, model, and output layers, with examples like feature attribution, knowledge graphs, and SHAP.
• An integrated end-to-end framework was synthesized to address opacity across all layers.
• Counterfactual approaches achieved the highest Explainability-Performance (E-P) score of 0.95.

Conclusions:

• The review offers a taxonomy, a cross-layer framework, and empirical data to guide the development of trustworthy PJRS.
• Future research should focus on multimodal causal inference, adaptive systems, and efficient explainability tools.
• This work supports the creation of more transparent and reliable AI-driven hiring systems.