Real or Robotic? Assessing Whether LLMs Accurately Simulate Qualities of Human Responses in Dialogue

TL;DR

This study evaluates whether large language models can accurately simulate human dialogue qualities by comparing generated LLM-LLM and human-LLM interactions, revealing significant divergence and language-specific performance patterns.

Contribution

The paper provides a large-scale analysis of LLM-generated dialogues versus human dialogues, highlighting the limitations and divergence in style and content across languages.

Findings

Low alignment between LLM simulations and human dialogues

Models perform similarly across English, Chinese, and Russian

LLMs better simulate human responses when writing style is similar to their own

Abstract

Studying and building datasets for dialogue tasks is both expensive and time-consuming due to the need to recruit, train, and collect data from study participants. In response, much recent work has sought to use large language models (LLMs) to simulate both human-human and human-LLM interactions, as they have been shown to generate convincingly human-like text in many settings. However, to what extent do LLM-based simulations \textit{actually} reflect human dialogues? In this work, we answer this question by generating a large-scale dataset of 100,000 paired LLM-LLM and human-LLM dialogues from the WildChat dataset and quantifying how well the LLM simulations align with their human counterparts. Overall, we find relatively low alignment between simulations and human interactions, demonstrating a systematic divergence along the multiple textual properties, including style and content. Further, in comparisons of English, Chinese, and Russian dialogues, we find that models perform similarly. Our results suggest that LLMs generally perform better when the human themself writes in a way that is more similar to the LLM's own style.