@inproceedings{grief-albert-etal-2026-emulate,
title = "Emulate or Estimate? The Divergent Strengths of Base and Post-Trained Language Models for Opinion Simulation",
author = "Grief-Albert, Seth and
Bo, Jessica and
Jiao, Difan and
Anderson, Ashton",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.1379/",
pages = "29311--29320",
ISBN = "979-8-89176-462-0",
abstract = "Large language models are increasingly used to simulate human opinions, but prior work reports conflicting results: some studies find promising alignment with human survey data, while others find persona collapse and weak demographic sensitivity. We propose that much of this conflict stems from conflating two distinct tasks. We call the first task emulation, in which models generate individual responses that aggregate into a population distribution. We call the second task estimation, in which models directly predict the population distribution. Evaluating six matched base and post-trained models on the Pew American Trends Panel, we find that base models are the stronger emulators: they produce response distributions closer to human ground truth and better preserve demographic structure. Post-trained models are generally the stronger estimators, producing more accurate distributional predictions when asked directly. We argue that model selection for human simulation should be guided by whether the task requires generating text or predicting distributions."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="grief-albert-etal-2026-emulate">
<titleInfo>
<title>Emulate or Estimate? The Divergent Strengths of Base and Post-Trained Language Models for Opinion Simulation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Seth</namePart>
<namePart type="family">Grief-Albert</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessica</namePart>
<namePart type="family">Bo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Difan</namePart>
<namePart type="family">Jiao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ashton</namePart>
<namePart type="family">Anderson</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Large language models are increasingly used to simulate human opinions, but prior work reports conflicting results: some studies find promising alignment with human survey data, while others find persona collapse and weak demographic sensitivity. We propose that much of this conflict stems from conflating two distinct tasks. We call the first task emulation, in which models generate individual responses that aggregate into a population distribution. We call the second task estimation, in which models directly predict the population distribution. Evaluating six matched base and post-trained models on the Pew American Trends Panel, we find that base models are the stronger emulators: they produce response distributions closer to human ground truth and better preserve demographic structure. Post-trained models are generally the stronger estimators, producing more accurate distributional predictions when asked directly. We argue that model selection for human simulation should be guided by whether the task requires generating text or predicting distributions.</abstract>
<identifier type="citekey">grief-albert-etal-2026-emulate</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.1379/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>29311</start>
<end>29320</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Emulate or Estimate? The Divergent Strengths of Base and Post-Trained Language Models for Opinion Simulation
%A Grief-Albert, Seth
%A Bo, Jessica
%A Jiao, Difan
%A Anderson, Ashton
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F grief-albert-etal-2026-emulate
%X Large language models are increasingly used to simulate human opinions, but prior work reports conflicting results: some studies find promising alignment with human survey data, while others find persona collapse and weak demographic sensitivity. We propose that much of this conflict stems from conflating two distinct tasks. We call the first task emulation, in which models generate individual responses that aggregate into a population distribution. We call the second task estimation, in which models directly predict the population distribution. Evaluating six matched base and post-trained models on the Pew American Trends Panel, we find that base models are the stronger emulators: they produce response distributions closer to human ground truth and better preserve demographic structure. Post-trained models are generally the stronger estimators, producing more accurate distributional predictions when asked directly. We argue that model selection for human simulation should be guided by whether the task requires generating text or predicting distributions.
%U https://aclanthology.org/2026.findings-emnlp.1379/
%P 29311-29320
Markdown (Informal)
[Emulate or Estimate? The Divergent Strengths of Base and Post-Trained Language Models for Opinion Simulation](https://aclanthology.org/2026.findings-emnlp.1379/) (Grief-Albert et al., Findings 2026)
ACL