DOI: 10.11118/978-80-7701-030-6-0126

GENERATIVNÍ UMĚLÁ INTELIGENCE (AI) JAKO ZRCADLO GENDEROVÝCH BIASŮ / GENERATIVE ARTIFICIAL INTELLIGENCE (AI) AS A MIRROR OF GENDER BIASES

Martin Richter ORCID...¹

¹ Fakulta sociálních věd, Univerzita Karlova, Smetanovo nábř. 6, 110 00 Praha – Staré Město

This article focuses on the issue of gender biases in generative artificial intelligence (AI) models, which can influence decision-making processes and the resulting societal dynamics. Through an analysis of responses from widely used language models (GPT-4o mini, GPT-4o, Llama 3.1, Claude 3.5 Sonnet, Google Gemini, and Mistral), this study investigates how these systems reproduce, amplify, or mitigate gender stereotypes in their generated outputs. By combining experimental design with content analysis, the model responses are classified as stereotypical, anti-stereotypical, or neutral. In this context, the results reveal significant differences between individual models. While some models, such as GPT-4o mini and Llama 3.1, systematically reproduce gender biases, others, like Claude 3.5 Sonnet and GPT-4o, exhibit a more inclusive approach. The study also highlights the significant social and economic implications of these biases, particularly in areas such as workforce recruitment and career counseling, where AI-generated biases can further exacerbate gender inequalities.

Keywords: Generative AI, LLMs, AI Generated Biases, ChatGPT

pages: 126-133, Published: 2025, online: 2025