Can LLMs Simulate L2-English Dialogue? An Information-Theoretic Analysis of L1-Dependent Biases

This paper investigates whether large language models can effectively simulate L2-English dialogue and analyzes L1-dependent biases using information-theoretic approaches. We examine how different first languages influence the quality and authenticity of simulated second-language conversations.

The research provides insights into the challenges of cross-linguistic dialogue simulation and contributes to understanding language transfer effects in AI systems.

Recommended citation: R Gao, X Wu, T Kuribayashi, M Ye, S Qi, C Roever, Y Liu, Z Yuan, JH Lau. (2025). "Can LLMs Simulate L2-English Dialogue? An Information-Theoretic Analysis of L1-Dependent Biases." Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025).
Download Paper