Personality Shapes Gender Bias in Persona-Conditioned LLM Narratives Across English and Hindi: An Empirical Investigation
arXiv:2604.23600v2 Announce Type: replace Abstract: Large Language Models (LLMs) are increasingly deployed in persona-driven applications such as education, customer service, and social platforms, where models are prompted to adopt specific personas when interacting with users. While persona conditioning can improve user experience and engagement, it also raises concerns about how personality cues may interact with gender biases and stereotypes. In this work, we present a controlled study of persona-conditioned story generation in English and Hindi, where each story portrays a working professi
The increasing deployment of LLMs in persona-driven applications makes understanding and mitigating inherent biases critical for their ethical and effective societal integration.
This research provides empirical evidence of how personality cues interact with gender biases in LLMs across languages, impacting the trustworthiness and fairness of AI interactions in diverse contexts.
We gain a deeper empirical understanding of the interplay between personality conditioning, gender bias, and cross-lingual differences in LLM narrative generation, highlighting challenges for responsible AI development.
- · AI ethicists
- · Developers of fairness-aware LLMs
- · Multilingual AI research
- · Unregulated LLM deployment
- · Generative AI applications with unmitigated bias
Increased focus on bias mitigation techniques specifically for persona-conditioned LLMs and for culturally diverse linguistic contexts.
Development of new regulatory frameworks or industry standards demanding bias detection and reduction in AI systems that generate personalized content.
Growing public distrust of AI systems that are perceived to perpetuate or amplify societal biases through personalized interactions.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.CL