September 20, 2025

Shaping Shared Languages: Human and Large Language Models' Inductive Biases in Emergent Communication

View Full Paper

Ask AI

Bookmark

Emergent artificial languages optimize communication through inductive biases, even in human-LLM collaborations.
Referentially grounded vocabularies develop across human-human, LLM-LLM, and human-LLM settings, showcasing their adaptability.
Human-LLM interactions lead to the emergence of vocabularies that resemble human-like communication more than LLM-like communication.
The study suggests that incorporating communicative success as a reward in LLM training can enhance alignment with human language.

Discussion

No takes yet. Share an insight, caveat, or question.

Research demonstrates how inductive biases influence language in human and large language models, revealing the importance of effective interaction.

Kouwenhoven et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark

Research demonstrates how inductive biases influence language in human and large language models, revealing the importance of effective interaction.