ABSTRACT The job applicant hiring process is inherently complex, encompassing various dimensions such as cultural fit, team dynamics, and individual qualifications. This paper explores the potential of machine learning techniques as analytical tools to examine the nuances of the hiring process rather than automating it. In this study, we use machine learning models to evaluate whether gender or age bias exists in hiring decisions by examining an extensive IT industry dataset comprising over 70,000 applicants. We provide practical guidelines for data preparation, bias evaluation, and assessment of AI-driven hiring outcomes, focusing on the influence of key factors on hiring probability. Our findings reveal that female applicants tend to receive lower hiring probabilities than male applicants, suggesting a gender-related bias. Additionally, we identify that proficiency in “Computer Skills,” the number of programming languages known, and familiarity with the language “TypeScript”—introduced in 2012—are among the top skills influencing hiring decisions. This study offers HR practitioners actionable insights into the AI decision-making process, highlighting how skills valued by AI models may differ from those prioritized in traditional recruitment. Through our empirical study, we demonstrate how multiple machine learning methods can be used for bias detection and the validation of critical decision factors in hiring decisions, equipping HR professionals with tools to enhance understanding and fairness in AI-driven hiring.
Read full abstract