ChatGPT Perpetuates Gender Bias in Machine Translation and Ignores Non-Gendered Pronouns: Findings across Bengali and Five other Low-Resource Languages

ChatGPT reinforces occupational gender stereotypes in translation: When translating gender-neutral pronouns from Bengali and five other languages, ChatGPT defaults to gendered pronouns based on stereotypes—translating neutral pronouns as "he" for doctors and "she" for nurses, perpetuating the assumption that certain professions are inherently male or female.

Gender-neutral language is being erased: ChatGPT completely fails to translate the English gender-neutral pronoun "they" into equivalent gender-neutral pronouns in other languages, instead producing incoherent and incorrect translations that eliminate inclusive language options.

The problem affects multiple languages globally: The bias was not limited to Bengali (the 7th most spoken language worldwide) but was consistent across Farsi, Malay, Tagalog, Thai, and Turkish—suggesting this is a systemic issue affecting translation between English and languages with gender-neutral pronouns.

ChatGPT shows a gender hierarchy even when given explicit instructions: While ChatGPT can produce gender-marked translations when prompted with gender information in English, the research found it "confers a higher respect to men than to women in the same occupation," indicating bias persists even with direct guidance.

This mirrors problems in earlier AI translation tools: ChatGPT exhibits the same gender biases previously documented in Google Translate and Microsoft Translator, suggesting the AI industry has not adequately addressed this known problem across generations of translation systems.