@inproceedings{wachter-etal-2025-llms,
title = "Are {LLM}s (Really) Ideological? An {IRT}-based Analysis and Alignment Tool for Perceived Socio-Economic Bias in {LLM}s",
author = "Wachter, Jasmin and
Radloff, Michael and
Smolej, Maja and
Kinder-Kurlanda, Katharina",
editor = "Arviv, Ofir and
Clinciu, Miruna and
Dhole, Kaustubh and
Dror, Rotem and
Gehrmann, Sebastian and
Habba, Eliya and
Itzhak, Itay and
Mille, Simon and
Perlitz, Yotam and
Santus, Enrico and
Sedoc, Jo{\~a}o and
Shmueli Scheuer, Michal and
Stanovsky, Gabriel and
Tafjord, Oyvind",
booktitle = "Proceedings of the Fourth Workshop on Generation, Evaluation and Metrics (GEM{\texttwosuperior})",
month = jul,
year = "2025",
address = "Vienna, Austria and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.gem-1.9/",
pages = "99--120",
ISBN = "979-8-89176-261-9",
abstract = "We introduce an Item Response Theory (IRT)-based framework to detect and quantify ideological bias in large language models (LLMs) without relying on subjective human judgments. Unlike prior work, our two-stage approach distinguishes between response avoidance and expressed bias by modeling `Prefer Not to Answer' (PNA) behaviors and calibrating ideological leanings based on open-ended responses. We fine-tune two LLM families to represent liberal and conservative baselines, and validate our approach using a 105-item ideological test inventory. Our results show that off-the-shelve LLMs frequently avoid engagement with ideological prompts, calling into question previous claims of partisan bias. This framework provides a statistically grounded and scalable tool for LLM alignment and fairness assessment. The general methodolody can also be applied to other forms of bias and languages."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wachter-etal-2025-llms">
<titleInfo>
<title>Are LLMs (Really) Ideological? An IRT-based Analysis and Alignment Tool for Perceived Socio-Economic Bias in LLMs</title>
</titleInfo>
<name type="personal">
<namePart type="given">Jasmin</namePart>
<namePart type="family">Wachter</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Michael</namePart>
<namePart type="family">Radloff</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Maja</namePart>
<namePart type="family">Smolej</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Katharina</namePart>
<namePart type="family">Kinder-Kurlanda</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Fourth Workshop on Generation, Evaluation and Metrics (GEM²)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Ofir</namePart>
<namePart type="family">Arviv</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Miruna</namePart>
<namePart type="family">Clinciu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kaustubh</namePart>
<namePart type="family">Dhole</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rotem</namePart>
<namePart type="family">Dror</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sebastian</namePart>
<namePart type="family">Gehrmann</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Eliya</namePart>
<namePart type="family">Habba</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Itay</namePart>
<namePart type="family">Itzhak</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Simon</namePart>
<namePart type="family">Mille</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yotam</namePart>
<namePart type="family">Perlitz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Enrico</namePart>
<namePart type="family">Santus</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">João</namePart>
<namePart type="family">Sedoc</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Michal</namePart>
<namePart type="family">Shmueli Scheuer</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gabriel</namePart>
<namePart type="family">Stanovsky</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Oyvind</namePart>
<namePart type="family">Tafjord</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Vienna, Austria and virtual meeting</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-261-9</identifier>
</relatedItem>
<abstract>We introduce an Item Response Theory (IRT)-based framework to detect and quantify ideological bias in large language models (LLMs) without relying on subjective human judgments. Unlike prior work, our two-stage approach distinguishes between response avoidance and expressed bias by modeling ‘Prefer Not to Answer’ (PNA) behaviors and calibrating ideological leanings based on open-ended responses. We fine-tune two LLM families to represent liberal and conservative baselines, and validate our approach using a 105-item ideological test inventory. Our results show that off-the-shelve LLMs frequently avoid engagement with ideological prompts, calling into question previous claims of partisan bias. This framework provides a statistically grounded and scalable tool for LLM alignment and fairness assessment. The general methodolody can also be applied to other forms of bias and languages.</abstract>
<identifier type="citekey">wachter-etal-2025-llms</identifier>
<location>
<url>https://aclanthology.org/2025.gem-1.9/</url>
</location>
<part>
<date>2025-07</date>
<extent unit="page">
<start>99</start>
<end>120</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Are LLMs (Really) Ideological? An IRT-based Analysis and Alignment Tool for Perceived Socio-Economic Bias in LLMs
%A Wachter, Jasmin
%A Radloff, Michael
%A Smolej, Maja
%A Kinder-Kurlanda, Katharina
%Y Arviv, Ofir
%Y Clinciu, Miruna
%Y Dhole, Kaustubh
%Y Dror, Rotem
%Y Gehrmann, Sebastian
%Y Habba, Eliya
%Y Itzhak, Itay
%Y Mille, Simon
%Y Perlitz, Yotam
%Y Santus, Enrico
%Y Sedoc, João
%Y Shmueli Scheuer, Michal
%Y Stanovsky, Gabriel
%Y Tafjord, Oyvind
%S Proceedings of the Fourth Workshop on Generation, Evaluation and Metrics (GEM²)
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria and virtual meeting
%@ 979-8-89176-261-9
%F wachter-etal-2025-llms
%X We introduce an Item Response Theory (IRT)-based framework to detect and quantify ideological bias in large language models (LLMs) without relying on subjective human judgments. Unlike prior work, our two-stage approach distinguishes between response avoidance and expressed bias by modeling ‘Prefer Not to Answer’ (PNA) behaviors and calibrating ideological leanings based on open-ended responses. We fine-tune two LLM families to represent liberal and conservative baselines, and validate our approach using a 105-item ideological test inventory. Our results show that off-the-shelve LLMs frequently avoid engagement with ideological prompts, calling into question previous claims of partisan bias. This framework provides a statistically grounded and scalable tool for LLM alignment and fairness assessment. The general methodolody can also be applied to other forms of bias and languages.
%U https://aclanthology.org/2025.gem-1.9/
%P 99-120
Markdown (Informal)
[Are LLMs (Really) Ideological? An IRT-based Analysis and Alignment Tool for Perceived Socio-Economic Bias in LLMs](https://aclanthology.org/2025.gem-1.9/) (Wachter et al., GEM 2025)
ACL