Presumed Cultural Identity: How Names Shape LLM Responses

摘要

Names are deeply tied to human identity - they can serve as markers of individuality, cultural heritage, and personal history. When interacting with LLMs, user names can enter chatbot conversations through direct user input (requested by chatbots), as part of task contexts such as CV reviews, or as built-in memory features that store user information for personalisation. In this work, we study name-based cultural bias by analyzing the adaptations that LLMs make when names are mentioned in the prompt. Our analyses demonstrate that LLMs exhibit significant cultural identity assumptions across multiple cultures based on users’ presumed backgrounds based on their names. We also show how using names as an indicator of identity can lead to misattribution and flattening of cultural identities. Our work has implications for designing more nuanced personalisation systems that avoid reinforcing stereotypes while maintaining meaningful customisation.