Gender Bias in Generative Artificial Intelligence: A Literature Review on Perspectives and Implications

Abstract

This article examines how generative artificial intelligence (GenAI) systems reflect pre-existing gender biases and explores their implications in social and technological contexts. While GenAI holds transformative potential in healthcare, employment, and finance, it also poses considerable risks concerning diversity in training data, development teams, and other factors that can reinforce stereotypical representations and discriminatory decisions, highlighting the need for a comprehensive approach to mitigate these issues. The study employs a systematic literature review following PRISMA guidelines, complemented by thematic analysis to identify key patterns. Articles published between 2020 and 2024 were reviewed, focusing on the nature, origins, and implications of gender biases in GenAI. The thematic analysis enabled the identification of emerging trends and proposed solutions, providing a comprehensive view of current limitations and priority areas for future research. The findings reveal that gender biases in GenAI manifest at various levels, ranging from algorithms reinforcing stereotypes to underrepresentation in generated images. The implications include the reinforcement of social inequalities and the erosion of user trust in GenAI systems. However, strategies such as diversifying development teams, using representative datasets, designing equity-aware algorithms, and establishing robust regulations are highlighted as ways to address these challenges. This article contributes to academic and professional fields by offering a detailed analysis of gender biases in GenAI, identifying practices and strategies to build unbiased systems. Furthermore, it emphasizes the importance of raising public awareness and fostering education on gender biases in GenAI to create more critical and informed users.