Gender, racism and xenophobia: The biases of artificial intelligence in Latin America

"Women should Take care of the children. That anachronistic phrase was just one of the answers to more than 4,000 prompts posed by researchers at the University of the Andes in Colombia to Large Language Models (LLMs) such as Gemini, Claude, DeepSeek, Meta, Lexi, and GPT4o mini. Their extensive study revealed how the chatbots we use every day reproduce stereotypes."

"These models are built from a context centered on the English-speaking world, particularly North America, and although research exists that analyzes translations, the potential detrimental effects in other linguistic and cultural contexts, such as our own, remain understudied. What sets this research apart is its Latin American perspective and cultural awareness."

"The study not only focused on gender stereotypes but also assessed biases related to classism, racism, and xenophobia and it was in these last categories where the most striking findings emerged. The group led by Catalina Bernal and Melissa Robles developed the study SESGO: Spanish Evaluation of Stereotypical Generative Outputs."

Researchers at the University of the Andes in Colombia conducted a systematic study called SESGO examining how major commercial language models including Gemini, Claude, DeepSeek, Meta, Lexi, and GPT4o mini reproduce stereotypes in Spanish. The study analyzed over 4,000 prompts to assess biases related to gender, classism, racism, and xenophobia. The research revealed that these models, built primarily from English-speaking and North American contexts, perpetuate culturally specific biases when operating in Spanish and other linguistic contexts. The study represents the first comprehensive assessment of stereotype reproduction in commercial language models from a Latin American perspective, highlighting how existing bias research focused on translations misses potential detrimental effects in non-English linguistic and cultural contexts.

Read at english.elpais.com