Characterizing Similarities and Divergences in Conversational Tones in Humans and LLMs by Sampling with People

TL;DR

This study introduces an iterative, cognitive science-inspired method to compare conversational tones in humans and GPT-4, revealing divergences and similarities through a large-scale, interpretable analysis of tone relations.

Contribution

It presents a novel iterative sampling approach for eliciting and analyzing conversational tones in humans and LLMs, reducing bias and improving representativeness.

Findings

Identified key differences and similarities in conversational tones between humans and GPT-4.

Created an interpretable geometric representation of tone relations.

Generated a large dataset of tone-labeled sentences for future research.

Abstract

Conversational tones -- the manners and attitudes in which speakers communicate -- are essential to effective communication. Amidst the increasing popularization of Large Language Models (LLMs) over recent years, it becomes necessary to characterize the divergences in their conversational tones relative to humans. However, existing investigations of conversational modalities rely on pre-existing taxonomies or text corpora, which suffer from experimenter bias and may not be representative of real-world distributions for the studies' psycholinguistic domains. Inspired by methods from cognitive science, we propose an iterative method for simultaneously eliciting conversational tones and sentences, where participants alternate between two tasks: (1) one participant identifies the tone of a given sentence and (2) a different participant generates a sentence based on that tone. We run 100 iterations of this process with human participants and GPT-4, then obtain a dataset of sentences and frequent conversational tones. In an additional experiment, humans and GPT-4 annotated all sentences with all tones. With data from 1,339 human participants, 33,370 human judgments, and 29,900 GPT-4 queries, we show how our approach can be used to create an interpretable geometric representation of relations between conversational tones in humans and GPT-4. This work demonstrates how combining ideas from machine learning and cognitive science can address challenges in human-computer interactions.