Gender Bias in AI: Examination of Contributing Factors and Mitigating Strategies

Abstract

Information systems (IS) research on the management of gender bias in AI and its negative repercussions, including techniques for reducing gender bias in AI, is lacking despite the information system (IS) field's recognition of the rich contribution of AI-based outcomes and their effects. Consequently, the growing issue of gender bias in AI is receiving more attention. More specifically, there is a need for a deeper comprehension on strategies for mitigating gender bias in AI.

Therefore, the objective of this research is: (i) To review and examine what gender bias entails in the context of AI use, (ii) Investigating and presenting the role of relevant contributing factors behind gender bias in AI and the approaches that could mitigate the gender bias in AI, (iii) Designing a conceptual framework to theorize gender bias in AI.

Accordingly, the research design of this research comprises three parts: (i) systematic literature review, (ii) interview with experts, (iii) Delphi study. To have a better understanding of gender bias in AI, this research is situated in the context of AI recruitment systems. Interpretivism was used as a philosophical foundation for the interview with experts as it enabled the explanation of the findings, offering conceptual insights, and providing a comprehensive narrative of the experts’ perspective. Lastly, a Delphi study was conducted to achieve consensus from experts for further refining and reliability of the findings and framework.

In this study a variety of research techniques were used to strengthen and improve its findings as well as to facilitate the triangulation approach. A theoretical framework that combines technical, organizational, and societal methods is suggested in light of the research's findings. Further, institutional theory has been used as a lens to tease out a better understanding of the mitigating approaches which are discussed through the mechanism of institutional isomorphism.

The empirical data gathered from interview with experts and the Delphi study provided unique insights on the research phenomenon. Based on the empirical findings, This study conceptualizes gender bias in AI as a socio-technical issue and offers a theoretical framework for its mitigation through socio-technical methods in an organizational setting. The theoretical framework offers practitioners and researchers an approach for examining the contributing factors of gender bias in AI and presents socio- technical approaches to inform management. Further, this study also explores the barriers and effects of managing gender bias in AI.