@inproceedings{francis-etal-2026-chatgpt,
title = "{C}hat{GPT}, why can{'}t anyone afford a house? On the Effects of {LLM} pre-annotation on Annotator Subjectivity",
author = "Francis, Emilie and
Leuzinger, C{\'e}line and
Mu{\~n}oz S{\'a}nchez, Ricardo and
Gauthier, Lee D.",
editor = "Dudy, Shiran and
Abercrombie, Gavin and
Basile, Valerio and
Leonardelli, Elisa and
Frenda, Simona",
booktitle = "Proceedings of the the fifth edition of {NLP}erspectives",
month = may,
year = "2026",
address = "Palma, Mallorca (Spain)",
publisher = "ELRA Language Resources Association (ELRA)",
url = "https://aclanthology.org/2026.nlperspectives-1.10/",
doi = "10.63317/4ab3uiguwubh",
pages = "98--111",
abstract = "Large language models (LLMs) have often been proposed as substitutes for human annotators in a variety of tasks. At the same time, there has been increased focus on the role that human subjectivity and perspective plays in data annotation. To avoid eliminating the human role in annotation entirely, the use of LLMs for pre-annotation has been suggested as an alternative approach. In this paper, we explore to which degree this approach affects subjectivity of social media annotation in English. We focus on comments regarding the current status of the housing market and label them for concern level, factors affecting housing affordability, and aspects that authors claim either exacerbate or improve the situation. To investigate this, we design an experiment involving two rounds of annotation: the first, a dataset annotated by humans only; and the second, a dataset with LLM pre-annotations curated by the same human annotators. We observe that the second setting leads to much higher agreement, as well as significant changes in label distribution and co-occurrence. Similar shifts do not appear in the LLM labels. Our findings show that use of LLMs in the annotation process leads to convergence in annotations and, thus, to an erosion of human subjectivity."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="francis-etal-2026-chatgpt">
<titleInfo>
<title>ChatGPT, why can’t anyone afford a house? On the Effects of LLM pre-annotation on Annotator Subjectivity</title>
</titleInfo>
<name type="personal">
<namePart type="given">Emilie</namePart>
<namePart type="family">Francis</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Céline</namePart>
<namePart type="family">Leuzinger</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ricardo</namePart>
<namePart type="family">Muñoz Sánchez</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lee</namePart>
<namePart type="given">D</namePart>
<namePart type="family">Gauthier</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-05</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the the fifth edition of NLPerspectives</title>
</titleInfo>
<name type="personal">
<namePart type="given">Shiran</namePart>
<namePart type="family">Dudy</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gavin</namePart>
<namePart type="family">Abercrombie</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Valerio</namePart>
<namePart type="family">Basile</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Elisa</namePart>
<namePart type="family">Leonardelli</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Simona</namePart>
<namePart type="family">Frenda</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>ELRA Language Resources Association (ELRA)</publisher>
<place>
<placeTerm type="text">Palma, Mallorca (Spain)</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Large language models (LLMs) have often been proposed as substitutes for human annotators in a variety of tasks. At the same time, there has been increased focus on the role that human subjectivity and perspective plays in data annotation. To avoid eliminating the human role in annotation entirely, the use of LLMs for pre-annotation has been suggested as an alternative approach. In this paper, we explore to which degree this approach affects subjectivity of social media annotation in English. We focus on comments regarding the current status of the housing market and label them for concern level, factors affecting housing affordability, and aspects that authors claim either exacerbate or improve the situation. To investigate this, we design an experiment involving two rounds of annotation: the first, a dataset annotated by humans only; and the second, a dataset with LLM pre-annotations curated by the same human annotators. We observe that the second setting leads to much higher agreement, as well as significant changes in label distribution and co-occurrence. Similar shifts do not appear in the LLM labels. Our findings show that use of LLMs in the annotation process leads to convergence in annotations and, thus, to an erosion of human subjectivity.</abstract>
<identifier type="citekey">francis-etal-2026-chatgpt</identifier>
<identifier type="doi">10.63317/4ab3uiguwubh</identifier>
<location>
<url>https://aclanthology.org/2026.nlperspectives-1.10/</url>
</location>
<part>
<date>2026-05</date>
<extent unit="page">
<start>98</start>
<end>111</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T ChatGPT, why can’t anyone afford a house? On the Effects of LLM pre-annotation on Annotator Subjectivity
%A Francis, Emilie
%A Leuzinger, Céline
%A Muñoz Sánchez, Ricardo
%A Gauthier, Lee D.
%Y Dudy, Shiran
%Y Abercrombie, Gavin
%Y Basile, Valerio
%Y Leonardelli, Elisa
%Y Frenda, Simona
%S Proceedings of the the fifth edition of NLPerspectives
%D 2026
%8 May
%I ELRA Language Resources Association (ELRA)
%C Palma, Mallorca (Spain)
%F francis-etal-2026-chatgpt
%X Large language models (LLMs) have often been proposed as substitutes for human annotators in a variety of tasks. At the same time, there has been increased focus on the role that human subjectivity and perspective plays in data annotation. To avoid eliminating the human role in annotation entirely, the use of LLMs for pre-annotation has been suggested as an alternative approach. In this paper, we explore to which degree this approach affects subjectivity of social media annotation in English. We focus on comments regarding the current status of the housing market and label them for concern level, factors affecting housing affordability, and aspects that authors claim either exacerbate or improve the situation. To investigate this, we design an experiment involving two rounds of annotation: the first, a dataset annotated by humans only; and the second, a dataset with LLM pre-annotations curated by the same human annotators. We observe that the second setting leads to much higher agreement, as well as significant changes in label distribution and co-occurrence. Similar shifts do not appear in the LLM labels. Our findings show that use of LLMs in the annotation process leads to convergence in annotations and, thus, to an erosion of human subjectivity.
%R 10.63317/4ab3uiguwubh
%U https://aclanthology.org/2026.nlperspectives-1.10/
%U https://doi.org/10.63317/4ab3uiguwubh
%P 98-111
Markdown (Informal)
[ChatGPT, why can’t anyone afford a house? On the Effects of LLM pre-annotation on Annotator Subjectivity](https://aclanthology.org/2026.nlperspectives-1.10/) (Francis et al., NLPerspectives 2026)
ACL