Multi-turn Evaluation of Anthropomorphic Behaviours in Large Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ibrahim, Lujain, Akbulut, Canfer, Elasmar, Rasmi, Rastogi, Charvi, Kahng, Minsuk, Morris, Meredith Ringel, McKee, Kevin R., Rieser, Verena, Shanahan, Murray, Weidinger, Laura
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910009187631104
author Ibrahim, Lujain
Akbulut, Canfer
Elasmar, Rasmi
Rastogi, Charvi
Kahng, Minsuk
Morris, Meredith Ringel
McKee, Kevin R.
Rieser, Verena
Shanahan, Murray
Weidinger, Laura
author_facet Ibrahim, Lujain
Akbulut, Canfer
Elasmar, Rasmi
Rastogi, Charvi
Kahng, Minsuk
Morris, Meredith Ringel
McKee, Kevin R.
Rieser, Verena
Shanahan, Murray
Weidinger, Laura
contents The tendency of users to anthropomorphise large language models (LLMs) is of growing interest to AI developers, researchers, and policy-makers. Here, we present a novel method for empirically evaluating anthropomorphic LLM behaviours in realistic and varied settings. Going beyond single-turn static benchmarks, we contribute three methodological advances in state-of-the-art (SOTA) LLM evaluation. First, we develop a multi-turn evaluation of 14 anthropomorphic behaviours. Second, we present a scalable, automated approach by employing simulations of user interactions. Third, we conduct an interactive, large-scale human subject study (N=1101) to validate that the model behaviours we measure predict real users' anthropomorphic perceptions. We find that all SOTA LLMs evaluated exhibit similar behaviours, characterised by relationship-building (e.g., empathy and validation) and first-person pronoun use, and that the majority of behaviours only first occur after multiple turns. Our work lays an empirical foundation for investigating how design choices influence anthropomorphic model behaviours and for progressing the ethical debate on the desirability of these behaviours. It also showcases the necessity of multi-turn evaluations for complex social phenomena in human-AI interaction.
format Preprint
id arxiv_https___arxiv_org_abs_2502_07077
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Multi-turn Evaluation of Anthropomorphic Behaviours in Large Language Models
Ibrahim, Lujain
Akbulut, Canfer
Elasmar, Rasmi
Rastogi, Charvi
Kahng, Minsuk
Morris, Meredith Ringel
McKee, Kevin R.
Rieser, Verena
Shanahan, Murray
Weidinger, Laura
Computation and Language
Computers and Society
Human-Computer Interaction
The tendency of users to anthropomorphise large language models (LLMs) is of growing interest to AI developers, researchers, and policy-makers. Here, we present a novel method for empirically evaluating anthropomorphic LLM behaviours in realistic and varied settings. Going beyond single-turn static benchmarks, we contribute three methodological advances in state-of-the-art (SOTA) LLM evaluation. First, we develop a multi-turn evaluation of 14 anthropomorphic behaviours. Second, we present a scalable, automated approach by employing simulations of user interactions. Third, we conduct an interactive, large-scale human subject study (N=1101) to validate that the model behaviours we measure predict real users' anthropomorphic perceptions. We find that all SOTA LLMs evaluated exhibit similar behaviours, characterised by relationship-building (e.g., empathy and validation) and first-person pronoun use, and that the majority of behaviours only first occur after multiple turns. Our work lays an empirical foundation for investigating how design choices influence anthropomorphic model behaviours and for progressing the ethical debate on the desirability of these behaviours. It also showcases the necessity of multi-turn evaluations for complex social phenomena in human-AI interaction.
title Multi-turn Evaluation of Anthropomorphic Behaviours in Large Language Models
topic Computation and Language
Computers and Society
Human-Computer Interaction
url https://arxiv.org/abs/2502.07077