As chatbots become integral to daily life, per-sonalizing systems is key for fostering trust, en-gagement, and inclusivity. This study examineshow linguistic similarity affects chatbot perfor-mance, focusing on integrating African Ameri-can English (AAE) into virtual agents to betterserve the African American community. Wedevelop text-based and spoken chatbots usinglarge language models and text-to-speech tech-nology, then evaluate them with AAE speakersagainst standard English chatbots. Our resultsshow that while text-based AAE chatbots of-ten underperform, spoken chatbots benefit froman African American voice and AAE elements,improving performance and preference. Thesefindings underscore the complexities of linguis-tic personalization and the dynamics betweentext and speech modalities, highlighting tech-nological limitations that affect chatbots’ AAspeech generation and pointing to promisingfuture research directions.
Proceedings of the Annual Meeting of the Association for Computational Linguistics (ACL) / 2025