Ethical AI for recruitment: a hybrid machine learning approach with fairness metrics and explainability tools

Abstract

Artificial Intelligence (AI) is transforming the process of recruitment making it efficient, consistent, and more accurate in its decisions; however, there are still ethical concerns like bias and obscurity. This paper constructs an Ethical AI Recruitment Framework and based on a hybrid machine learning model that incorporates a combination of Logistic Regression, Decision Tree, and Random Forest with soft voting to achieve a balance between predictive accuracy and fairness and the final accuracy was measured 97.4% using the hybrid model. Demographic Parity (DP), Equalized Odds (EO), and a Gender-Flip Test were used to evaluate the system and a +5% threshold recalibration was applied to minimize residual bias. The findings demonstrate high technical and ethical results with an accuracy of 96% and a Fairness Index of 0.87, Transparency Score of 0.01 and Bias Volatility of 0.027, which ensures reliability and fairness. Explainability SHAP and LIME demonstrated that skills, education and experience were the most influential features in hiring decisions and that gender and age became insignificant after applying these features. The Selection Scoreboard (80% model prediction and 20% resumes strength) of the framework offers a clear ranking of the candidates. In general, this study shows that fairness, transparency, and accountability can be appropriately incorporated into AIbased recruitment to provide a repeatable model of reliable and bias-resistant decision making in the HR field.