The Role of Artificial Intelligence (AI) in Resume Screening
Abidizadegan, Fatemeh (2025)
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2025120934179
https://urn.fi/URN:NBN:fi:amk-2025120934179
Tiivistelmä
This thesis investigates the ethical dimensions and societal implications of utilizing AI for the automatic filtering of resumes, a practice that has seen rapid growth within the recruitment industry. Driven by this trend, organizations are increasingly employing machine learning and large language models (LLMs) on a large scale for candidate assessment. Consequently, concerns regarding bias, fairness, transparency, and accountability have gained prominence. This research, through a comprehensive review of 40 peer-reviewed articles published between 2015 and 2025, identifies various types of bias documented in the literature.
The analysis focuses on discrimination related to gender, race, age, socioeconomic status, and the discriminatory use of names. It also pinpoints the origins of these biases in the training data, algorithmic design, and the decision-making processes of organizations. This thesis evaluates several fairness metrics, including statistical parity, equal opportunity, and predictive parity. Additionally, it considers moral and legal frameworks such as the EU AI Act, GDPR, IEEE standards, and OECD principles. While AI has the potential to enhance efficiency and consistency in resume screening, the findings of this report indicate that current AI systems largely fail to adequately address transparency issues and often perpetuate historical patterns of discrimination.
This survey highlights the gap between local requirements, academic research, and industry practices, suggesting that operational implementations are rarely assessed and that there is a significant absence of standard fairness benchmarks. The core concepts of the anti-bias initiative include depoliticizing algorithms, ensuring explainability in AI, conducting audits, and incorporating human oversight.
The research posits that AI tools could be utilized to create more equitable and efficient hiring processes if measures such as governance, transparency, and ongoing human supervision are established. The author emphasizes the significance of interdisciplinary research and the ethical application of such research in this context.
The analysis focuses on discrimination related to gender, race, age, socioeconomic status, and the discriminatory use of names. It also pinpoints the origins of these biases in the training data, algorithmic design, and the decision-making processes of organizations. This thesis evaluates several fairness metrics, including statistical parity, equal opportunity, and predictive parity. Additionally, it considers moral and legal frameworks such as the EU AI Act, GDPR, IEEE standards, and OECD principles. While AI has the potential to enhance efficiency and consistency in resume screening, the findings of this report indicate that current AI systems largely fail to adequately address transparency issues and often perpetuate historical patterns of discrimination.
This survey highlights the gap between local requirements, academic research, and industry practices, suggesting that operational implementations are rarely assessed and that there is a significant absence of standard fairness benchmarks. The core concepts of the anti-bias initiative include depoliticizing algorithms, ensuring explainability in AI, conducting audits, and incorporating human oversight.
The research posits that AI tools could be utilized to create more equitable and efficient hiring processes if measures such as governance, transparency, and ongoing human supervision are established. The author emphasizes the significance of interdisciplinary research and the ethical application of such research in this context.