ALGORITHMIC BIAS AND SDG EQUITY: ISSUES WITH REPRESENTATION IN AI OUTPUTS
UKLO Proceedings · 09.05.2026 · 2 (1) · 251-260
DOI: 10.20544/uklop.2026.630 ↗Kurz gefasst
Fachartikel in der Zeitschrift UKLO Proceedings, online erschienen am 09.05.2026. Verfasst von Elena Shalevska und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
AI is now everywhere, shaping how we search, write, and even create images. Recognizing this, this study investigates gender stereotypes in the image outputs of five leading generative artificial intelligence (AI) platforms such as: Grok, Gemini (Nano Banana), ChatGPT (DALL·E), and DeepSeek. Using a small set of prompts that focus on occupations (e.g., team coach, judge, housekeeper, computer programmer, primary school teacher), this qualitative research examines whether AI systems reproduce or challenge existing societal biases. Initial findings indicate that the generated images frequently reflect traditional gender norms: high-status roles such as judge or coach are predominantly depicted as male, while service-oriented or care-related roles, such as cleaning staff and primary school teacher, are more often portrayed as female. These results suggest that AI models, trained on large-scale human data, tend to mimic rather than mitigate human gender bias, posing challenges for Sustainable Development Goals (SDG) 5 (Gender Equality) and 10 (Reduced Inequalities).
Zitieren
APA
Shalevska, E., & Kostadinovska-Stojchevska, B. (2026). ALGORITHMIC BIAS AND SDG EQUITY: ISSUES WITH REPRESENTATION IN AI OUTPUTS. UKLO Proceedings, 2(1), 251-260. https://doi.org/10.20544/uklop.2026.630
Vancouver
Shalevska E, Kostadinovska-Stojchevska B. ALGORITHMIC BIAS AND SDG EQUITY: ISSUES WITH REPRESENTATION IN AI OUTPUTS. UKLO Proceedings. 2026;2(1):251-260. doi:10.20544/uklop.2026.630
BibTeX
@article{Shalevska2026,
author = {Shalevska, Elena and Kostadinovska-Stojchevska, Bisera},
title = {ALGORITHMIC BIAS AND SDG EQUITY: ISSUES WITH REPRESENTATION IN AI OUTPUTS},
journal = {UKLO Proceedings},
year = {2026},
volume = {2},
number = {1},
pages = {251--260},
doi = {10.20544/uklop.2026.630}
}
EUVIMED