Cognitive phantoms in large language models through the lens of latent variables

Large language models (LLMs) increasingly reach real-world applications, necessitating a better understanding of their behaviour. Their size and complexity complicate traditional assessment methods, causing the emergence of alternative approaches inspired by the field of psychology. Recent studies a...

Full description

Saved in:
Bibliographic Details
Main Authors: Sanne Peereboom, Inga Schwabe, Bennett Kleinberg
Format: Article
Language:English
Published: Elsevier 2025-05-01
Series:Computers in Human Behavior: Artificial Humans
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2949882125000453
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850246198115434496
author Sanne Peereboom
Inga Schwabe
Bennett Kleinberg
author_facet Sanne Peereboom
Inga Schwabe
Bennett Kleinberg
author_sort Sanne Peereboom
collection DOAJ
description Large language models (LLMs) increasingly reach real-world applications, necessitating a better understanding of their behaviour. Their size and complexity complicate traditional assessment methods, causing the emergence of alternative approaches inspired by the field of psychology. Recent studies administering psychometric questionnaires to LLMs report human-like traits in LLMs, potentially influencing LLM behaviour. However, this approach suffers from a validity problem: it presupposes that these traits exist in LLMs and that they are measurable with tools designed for humans. Typical procedures rarely acknowledge the validity problem in LLMs, comparing and interpreting average LLM scores. This study investigates this problem by comparing latent structures of personality between humans and three LLMs using two validated personality questionnaires. Findings suggest that questionnaires designed for humans do not validly measure similar constructs in LLMs, and that these constructs may not exist in LLMs at all, highlighting the need for psychometric analyses of LLM responses to avoid chasing cognitive phantoms.
format Article
id doaj-art-e397ed41e7fb4d6f990bba6402fb6fbd
institution OA Journals
issn 2949-8821
language English
publishDate 2025-05-01
publisher Elsevier
record_format Article
series Computers in Human Behavior: Artificial Humans
spelling doaj-art-e397ed41e7fb4d6f990bba6402fb6fbd2025-08-20T01:59:14ZengElsevierComputers in Human Behavior: Artificial Humans2949-88212025-05-01410016110.1016/j.chbah.2025.100161Cognitive phantoms in large language models through the lens of latent variablesSanne Peereboom0Inga Schwabe1Bennett Kleinberg2Tilburg University, Department of Methodology and Statistics, Professor Cobbenhagenlaan 125, Tilburg, 5037 DB, the Netherlands; Corresponding author.Tilburg University, Department of Methodology and Statistics, Professor Cobbenhagenlaan 125, Tilburg, 5037 DB, the NetherlandsTilburg University, Department of Methodology and Statistics, Professor Cobbenhagenlaan 125, Tilburg, 5037 DB, the Netherlands; University College London, Department of Security and Crime Science, 35 Tavistock Square, London, WC1H 9EZ, England, UKLarge language models (LLMs) increasingly reach real-world applications, necessitating a better understanding of their behaviour. Their size and complexity complicate traditional assessment methods, causing the emergence of alternative approaches inspired by the field of psychology. Recent studies administering psychometric questionnaires to LLMs report human-like traits in LLMs, potentially influencing LLM behaviour. However, this approach suffers from a validity problem: it presupposes that these traits exist in LLMs and that they are measurable with tools designed for humans. Typical procedures rarely acknowledge the validity problem in LLMs, comparing and interpreting average LLM scores. This study investigates this problem by comparing latent structures of personality between humans and three LLMs using two validated personality questionnaires. Findings suggest that questionnaires designed for humans do not validly measure similar constructs in LLMs, and that these constructs may not exist in LLMs at all, highlighting the need for psychometric analyses of LLM responses to avoid chasing cognitive phantoms.http://www.sciencedirect.com/science/article/pii/S2949882125000453Large language modelsPsychometricsMachine behaviourLatent variable modellingValidity
spellingShingle Sanne Peereboom
Inga Schwabe
Bennett Kleinberg
Cognitive phantoms in large language models through the lens of latent variables
Computers in Human Behavior: Artificial Humans
Large language models
Psychometrics
Machine behaviour
Latent variable modelling
Validity
title Cognitive phantoms in large language models through the lens of latent variables
title_full Cognitive phantoms in large language models through the lens of latent variables
title_fullStr Cognitive phantoms in large language models through the lens of latent variables
title_full_unstemmed Cognitive phantoms in large language models through the lens of latent variables
title_short Cognitive phantoms in large language models through the lens of latent variables
title_sort cognitive phantoms in large language models through the lens of latent variables
topic Large language models
Psychometrics
Machine behaviour
Latent variable modelling
Validity
url http://www.sciencedirect.com/science/article/pii/S2949882125000453
work_keys_str_mv AT sannepeereboom cognitivephantomsinlargelanguagemodelsthroughthelensoflatentvariables
AT ingaschwabe cognitivephantomsinlargelanguagemodelsthroughthelensoflatentvariables
AT bennettkleinberg cognitivephantomsinlargelanguagemodelsthroughthelensoflatentvariables