Algorithmic Fairness in AI Recruitment Systems: Detecting and Correcting Bias in Resume Screening

Venkatapathy Balaji, Kanisha Pathy (2025) Algorithmic Fairness in AI Recruitment Systems: Detecting and Correcting Bias in Resume Screening. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (1MB) | Preview

Abstract

The applicant tracking system (ATS), many times interpreted as a "black box," is a prevalent technology used in recruitment at mass market organizations, providing no feedback to the applicant, and presents a risk of unintended bias. This project creates a prototype of an explainable ATS that is designed to have fairness as a key concern and be able to rank resumes as strong or weak using structured features extracted from the resume text and give suggestions for improvement. Given that the dataset does not contain actual hiring data, I used K-Means clustering to create meaningful strong or weak labels using ATS metrics (skills, education, certifications, designations, word count, composite ATS score). The Tab Transformer model was built and trained using engineered features and categorical contexts for each platform (LinkedIn, Naukri, and Indeed) and job categories (IT, HR, Healthcare, and Finance). The Tab Transformer model evaluates fairness based on non-sensitive groups relevant to ATS, for example, by platform and job category. Evaluation of fairness is performed using selection rates, disparate impact ratios, and equal opportunity metrics. Model predictions are generated for each resume and can be visualized using SHAP explanations to show how features positively and negatively contribute to the model's predictions. The full pipeline is available in GitHub: Resume-ATS-Screening-Fairness-App and deployed as an interactive Explainable ATS Resume Screening · Streamlit. It offered resume evaluation with SHAP visualizations, explanations of why each candidate was rejected with recommendations for improvement, fairness dashboards, and resume comparison modules. Thus, it offers a practical foundation for candidate centric, fair, and transparent ATS design.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Yaqoob, Abid UNSPECIFIED
Uncontrolled Keywords:	Algorithmic fairness; resume screening; bias-mitigation; ATS; SHAP; recruitment AI; transparency
Subjects:	Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence H Social Sciences > HD Industries. Land use. Labor > HD28 Management. Industrial Management > Human Resource Management Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning H Social Sciences > HD Industries. Land use. Labor > HD28 Management. Industrial Management > Human Resource Management > Recruitment
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	09 Sep 2026 11:02
Last Modified:	09 Sep 2026 11:02
URI:	https://norma.ncirl.ie/id/eprint/9925

Actions (login required)

View Item