CHI workshop paper on leveraging LLMs to design conversational agents for young children

Abstract: The development of conversational agents (CAs) for young children has been proven to enrich early childhood learning in a wide range of domains, such as language, math, and science. Given the affordances of large language models (LLMs), we propose several design considerations for LLM-powered CAs to further enhance children’s learning by supporting in-depth conversations, less-structured learning activities, and tailored content. We also point out the potential risks and limitations that may arise from LLMs. These include the risk of perpetuating bias and stereotypes present in the data used for training, as well as the limited interpretability of LLM-generated content.

He, K., Xu, Y., & Warschauer, M. (2023, Febuary). Leveraging Large Language Models to Improve the Design of Conversational Agents for Young Children’s Learning. CHI 2023 Workshop on Child-centred AI Design: Definition, Operation and Considerations, Hamburg, Germany.