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

2025-03-06

AI Summary

This study investigates how human and large language model (LLM) inductive biases jointly shape artificial language evolution, particularly examining how human–LLM collaboration influences linguistic dynamics and cross-agent communicative consistency. Method: Adopting the referential game paradigm, we systematically compare three interaction regimes—human–human, LLM–LLM, and human–LLM collaboration—using behavioral experiments, cross-agent interaction evaluation, lexical emergence analysis, and bias quantification. Contribution/Results: We first demonstrate that human–LLM collaboration significantly mitigates LLMs’ tendency to diverge from human linguistic patterns, yielding more human-like shared lexicons. Building on this, we propose a novel training paradigm that uses “communicative success” as a reward signal. Empirical results confirm: (i) robust, referentially grounded language systems consistently emerge across all conditions; (ii) human–LLM languages exhibit structural and distributional properties closest to human languages; and (iii) communicative success serves as an effective and feasible alignment-driving signal for cross-agent language learning.

Technology Category

• Natural Language Processing: (Large) Language Models
• Humans and AI: Learning Human Values and Preferences
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Languages are shaped by the inductive biases of their users. Using a classical referential game, we investigate how artificial languages evolve when optimised for inductive biases in humans and large language models (LLMs) via Human-Human, LLM-LLM and Human-LLM experiments. We show that referentially grounded vocabularies emerge that enable reliable communication in all conditions, even when humans and LLMs collaborate. Comparisons between conditions reveal that languages optimised for LLMs subtly differ from those optimised for humans. Interestingly, interactions between humans and LLMs alleviate these differences and result in vocabularies which are more human-like than LLM-like. These findings advance our understanding of how inductive biases in LLMs play a role in the dynamic nature of human language and contribute to maintaining alignment in human and machine communication. In particular, our work underscores the need to think of new methods that include human interaction in the training processes of LLMs, and shows that using communicative success as a reward signal can be a fruitful, novel direction.

Problem

• Investigates how artificial languages evolve with human and LLM inductive biases.
• Compares language differences optimized for humans versus LLMs.
• Explores human-LLM interaction effects on language alignment and communication.

Innovation

• Referential game optimizes human-LLM communication.
• Human-LLM interaction creates human-like vocabularies.
• Communicative success as reward in LLM training.