Fairness and Bias Detection in Artificial Intelligence-Driven Decision-Making in Employment Processes using Machine Learning Model

Abstract

This study investigated fairness and bias detection in Artificial Intelligence (AI)-driven employment decision-making using machine learning models. The research aimed to determine whether AI-based recruitment systems make fair hiring decisions by evaluating key fairness. Additionally, it conducted an in-depth study on the application of AI in hiring, focusing on how it might save hiring expenses by automating the screening of numerous applicants. A machine learning-based model was developed and implemented using Python to assess fairness and bias in AI recruitment processes. The study utilized a dataset of 1,000 applicant records containing demographic and professional attributes such as gender, education, skills, and years of experience. Random Forest (RForest), Gradient Boosting Machine (GBoosting), and Support Vector Machine (SVector) algorithms were evaluated based on predictive performance and fairness. The outcome of the work shows that the SVector achieved the highest classification accuracy of 87.0% and demonstrated the best overall fairness performance across all evaluated metrics. The SVector model exhibited lower levels of bias relating to gender, location, experience, and disability inclusion. Feature importance analysis showed that years of experience was the most influential factor in hiring decisions across all models, while the influence of gender varied between models. This work examined the importance of balancing predictive accuracy with fairness when deploying AI systems in recruitment. Based on the results, the SVector model was identified as the most suitable approach due to its strong performance and reduced bias. To further improve fairness in AI recruitment systems, the study recommends continuous monitoring and regular fairness audits. These measures can help ensure more ethical, transparent, and equitable AI-driven hiring practices.