@inproceedings{liu-etal-2026-illusion,
title = "Illusion of Alignment: Detecting Hidden Disagreement in Collaborative Dialogue",
author = "Liu, Kaiming and
Luo, Fuwen and
Wang, Ziyue and
Ju, Jinrui and
Liu, Yuxuan and
Lei, Xuanyu and
Lai, Yunghwei and
Li, Peng and
Liu, Yang",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.930/",
pages = "21395--21425",
ISBN = "979-8-89176-459-0",
abstract = "Collaborative dialogue can end with apparent agreement while participants still differ on goals, assumptions, or execution plans, creating an **illusion of alignment (IoA)**. A real-user study across 18 meetings confirms that IoA arises routinely in human collaboration. Yet IoA poses a paradox: if participants were aware of such disagreements, they would already be explicit; if not, they cannot articulate them when asked, leaving IoA invisible to both participants and observers. In this work, we make IoA detectable by generating diagnostic multiple-choice questions whose divergent answers across participants provide direct behavioral evidence of hidden disagreement. We construct **IoA-Suite**, a dataset and evaluation protocol for detecting hidden disagreement, spanning five task types and six domains. We find that even the best model attains only 49.5{\%} F1, with the bottleneck traced to private context that the dialogue does not surface. We then train **IoA-Prober-8B** based on IoA-Suite, reaching 51.8{\%} F1 on IoA-Suite. Across the aforementioned 18 real meetings, it surfaces 2.89 hidden disagreements per meeting that participants confirm they had not voiced, transferring to live human dialogue. Further, in multi-agent collaboration, pairing IoA-Prober-8B with LLM agents improves downstream task performance on BigCodeBench-Hard and HiddenBench."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="liu-etal-2026-illusion">
<titleInfo>
<title>Illusion of Alignment: Detecting Hidden Disagreement in Collaborative Dialogue</title>
</titleInfo>
<name type="personal">
<namePart type="given">Kaiming</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Fuwen</namePart>
<namePart type="family">Luo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ziyue</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jinrui</namePart>
<namePart type="family">Ju</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuxuan</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xuanyu</namePart>
<namePart type="family">Lei</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yunghwei</namePart>
<namePart type="family">Lai</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Peng</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yang</namePart>
<namePart type="family">Liu</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>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</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">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Collaborative dialogue can end with apparent agreement while participants still differ on goals, assumptions, or execution plans, creating an **illusion of alignment (IoA)**. A real-user study across 18 meetings confirms that IoA arises routinely in human collaboration. Yet IoA poses a paradox: if participants were aware of such disagreements, they would already be explicit; if not, they cannot articulate them when asked, leaving IoA invisible to both participants and observers. In this work, we make IoA detectable by generating diagnostic multiple-choice questions whose divergent answers across participants provide direct behavioral evidence of hidden disagreement. We construct **IoA-Suite**, a dataset and evaluation protocol for detecting hidden disagreement, spanning five task types and six domains. We find that even the best model attains only 49.5% F1, with the bottleneck traced to private context that the dialogue does not surface. We then train **IoA-Prober-8B** based on IoA-Suite, reaching 51.8% F1 on IoA-Suite. Across the aforementioned 18 real meetings, it surfaces 2.89 hidden disagreements per meeting that participants confirm they had not voiced, transferring to live human dialogue. Further, in multi-agent collaboration, pairing IoA-Prober-8B with LLM agents improves downstream task performance on BigCodeBench-Hard and HiddenBench.</abstract>
<identifier type="citekey">liu-etal-2026-illusion</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.930/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>21395</start>
<end>21425</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Illusion of Alignment: Detecting Hidden Disagreement in Collaborative Dialogue
%A Liu, Kaiming
%A Luo, Fuwen
%A Wang, Ziyue
%A Ju, Jinrui
%A Liu, Yuxuan
%A Lei, Xuanyu
%A Lai, Yunghwei
%A Li, Peng
%A Liu, Yang
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F liu-etal-2026-illusion
%X Collaborative dialogue can end with apparent agreement while participants still differ on goals, assumptions, or execution plans, creating an **illusion of alignment (IoA)**. A real-user study across 18 meetings confirms that IoA arises routinely in human collaboration. Yet IoA poses a paradox: if participants were aware of such disagreements, they would already be explicit; if not, they cannot articulate them when asked, leaving IoA invisible to both participants and observers. In this work, we make IoA detectable by generating diagnostic multiple-choice questions whose divergent answers across participants provide direct behavioral evidence of hidden disagreement. We construct **IoA-Suite**, a dataset and evaluation protocol for detecting hidden disagreement, spanning five task types and six domains. We find that even the best model attains only 49.5% F1, with the bottleneck traced to private context that the dialogue does not surface. We then train **IoA-Prober-8B** based on IoA-Suite, reaching 51.8% F1 on IoA-Suite. Across the aforementioned 18 real meetings, it surfaces 2.89 hidden disagreements per meeting that participants confirm they had not voiced, transferring to live human dialogue. Further, in multi-agent collaboration, pairing IoA-Prober-8B with LLM agents improves downstream task performance on BigCodeBench-Hard and HiddenBench.
%U https://aclanthology.org/2026.emnlp-main.930/
%P 21395-21425
Markdown (Informal)
[Illusion of Alignment: Detecting Hidden Disagreement in Collaborative Dialogue](https://aclanthology.org/2026.emnlp-main.930/) (Liu et al., EMNLP 2026)
ACL
- Kaiming Liu, Fuwen Luo, Ziyue Wang, Jinrui Ju, Yuxuan Liu, Xuanyu Lei, Yunghwei Lai, Peng Li, and Yang Liu. 2026. Illusion of Alignment: Detecting Hidden Disagreement in Collaborative Dialogue. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 21395–21425, Budapest, Hungary. Association for Computational Linguistics.