A publicly available benchmark for assessing large language models’ ability to predict how humans balance self-interest and the interest of others

Abstract Large language models (LLMs) hold enormous potential to assist humans in decision-making processes, from everyday to high-stake scenarios. However, as many human decisions carry social implications, for LLMs to be reliable assistants a necessary prerequisite is that they are able to capture...

Full description

Saved in:
Bibliographic Details
Main Authors: Valerio Capraro, Roberto Di Paolo, Veronica Pizziol
Format: Article
Language:English
Published: Nature Portfolio 2025-07-01
Series:Scientific Reports
Subjects:
Online Access:https://doi.org/10.1038/s41598-025-01715-7
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849334354140987392
author Valerio Capraro
Roberto Di Paolo
Veronica Pizziol
author_facet Valerio Capraro
Roberto Di Paolo
Veronica Pizziol
author_sort Valerio Capraro
collection DOAJ
description Abstract Large language models (LLMs) hold enormous potential to assist humans in decision-making processes, from everyday to high-stake scenarios. However, as many human decisions carry social implications, for LLMs to be reliable assistants a necessary prerequisite is that they are able to capture how humans balance self-interest and the interest of others. Here we introduce a novel, publicly available, benchmark to test LLM’s ability to predict how humans balance monetary self-interest and the interest of others. This benchmark consists of 106 textual instructions from dictator games experiments conducted with human participants from 12 countries, alongside with a compendium of actual human behavior in each experiment. We investigate the ability of four advanced chatbots against this benchmark. We find that none of these chatbots meet the benchmark. In particular, only GPT-4 and GPT-4o (not Bard nor Bing) correctly capture qualitative behavioral patterns, identifying three major classes of behavior: self-interested, inequity-averse, and fully altruistic. Nonetheless, GPT-4 and GPT-4o consistently underestimate self-interest, while overestimating altruistic behavior. In sum, this article introduces a publicly available resource for testing the capacity of LLMs to estimate human other-regarding preferences in economic decisions and reveals an “optimistic bias” in current versions of GPT.
format Article
id doaj-art-e1d1ca36370443169d604d82fb1b6357
institution Kabale University
issn 2045-2322
language English
publishDate 2025-07-01
publisher Nature Portfolio
record_format Article
series Scientific Reports
spelling doaj-art-e1d1ca36370443169d604d82fb1b63572025-08-20T03:45:35ZengNature PortfolioScientific Reports2045-23222025-07-0115111110.1038/s41598-025-01715-7A publicly available benchmark for assessing large language models’ ability to predict how humans balance self-interest and the interest of othersValerio Capraro0Roberto Di Paolo1Veronica Pizziol2Department of Psychology, University of Milan BicoccaDepartment of Economics and Management, University of ParmaDepartment of Economics, University of BolognaAbstract Large language models (LLMs) hold enormous potential to assist humans in decision-making processes, from everyday to high-stake scenarios. However, as many human decisions carry social implications, for LLMs to be reliable assistants a necessary prerequisite is that they are able to capture how humans balance self-interest and the interest of others. Here we introduce a novel, publicly available, benchmark to test LLM’s ability to predict how humans balance monetary self-interest and the interest of others. This benchmark consists of 106 textual instructions from dictator games experiments conducted with human participants from 12 countries, alongside with a compendium of actual human behavior in each experiment. We investigate the ability of four advanced chatbots against this benchmark. We find that none of these chatbots meet the benchmark. In particular, only GPT-4 and GPT-4o (not Bard nor Bing) correctly capture qualitative behavioral patterns, identifying three major classes of behavior: self-interested, inequity-averse, and fully altruistic. Nonetheless, GPT-4 and GPT-4o consistently underestimate self-interest, while overestimating altruistic behavior. In sum, this article introduces a publicly available resource for testing the capacity of LLMs to estimate human other-regarding preferences in economic decisions and reveals an “optimistic bias” in current versions of GPT.https://doi.org/10.1038/s41598-025-01715-7Generative artificial intelligenceHuman behaviorEconomic gamesDictator gameAltruism
spellingShingle Valerio Capraro
Roberto Di Paolo
Veronica Pizziol
A publicly available benchmark for assessing large language models’ ability to predict how humans balance self-interest and the interest of others
Scientific Reports
Generative artificial intelligence
Human behavior
Economic games
Dictator game
Altruism
title A publicly available benchmark for assessing large language models’ ability to predict how humans balance self-interest and the interest of others
title_full A publicly available benchmark for assessing large language models’ ability to predict how humans balance self-interest and the interest of others
title_fullStr A publicly available benchmark for assessing large language models’ ability to predict how humans balance self-interest and the interest of others
title_full_unstemmed A publicly available benchmark for assessing large language models’ ability to predict how humans balance self-interest and the interest of others
title_short A publicly available benchmark for assessing large language models’ ability to predict how humans balance self-interest and the interest of others
title_sort publicly available benchmark for assessing large language models ability to predict how humans balance self interest and the interest of others
topic Generative artificial intelligence
Human behavior
Economic games
Dictator game
Altruism
url https://doi.org/10.1038/s41598-025-01715-7
work_keys_str_mv AT valeriocapraro apubliclyavailablebenchmarkforassessinglargelanguagemodelsabilitytopredicthowhumansbalanceselfinterestandtheinterestofothers
AT robertodipaolo apubliclyavailablebenchmarkforassessinglargelanguagemodelsabilitytopredicthowhumansbalanceselfinterestandtheinterestofothers
AT veronicapizziol apubliclyavailablebenchmarkforassessinglargelanguagemodelsabilitytopredicthowhumansbalanceselfinterestandtheinterestofothers
AT valeriocapraro publiclyavailablebenchmarkforassessinglargelanguagemodelsabilitytopredicthowhumansbalanceselfinterestandtheinterestofothers
AT robertodipaolo publiclyavailablebenchmarkforassessinglargelanguagemodelsabilitytopredicthowhumansbalanceselfinterestandtheinterestofothers
AT veronicapizziol publiclyavailablebenchmarkforassessinglargelanguagemodelsabilitytopredicthowhumansbalanceselfinterestandtheinterestofothers