Illusion of Alignment: Detecting Hidden Disagreement in Collaborative Dialogue

2026-08-08

AI Summary

This work addresses the “illusion of alignment” (IoA)—a pervasive yet subtle phenomenon in collaborative dialogue where interlocutors appear aligned on the surface but harbor hidden divergences in goals, assumptions, or plans. To detect such latent misalignments, the authors propose a novel method that generates diagnostic multiple-choice questions and leverages discrepancies in participants’ answers as behavioral evidence of unspoken disagreements. They introduce IoA-Suite, a benchmark dataset spanning five task types and six domains, along with a dedicated evaluation protocol, and develop IoA-Prober-8B, a specialized large language model detector. Evaluated on IoA-Suite, the approach achieves an F1 score of 51.8%, uncovers an average of 2.89 hidden divergences per real-world meeting, and significantly enhances multi-agent collaborative performance on BigCodeBench-Hard and HiddenBench.

Abstract

Collaborative dialogue can end with apparent agreement while participants still differ on goals, assumptions, or execution plans, creating an \textbf{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 \textbf{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 \textbf{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.

Problem

• illusion of alignment
• hidden disagreement
• collaborative dialogue
• alignment detection
• dialogue analysis

Innovation

• Illusion of Alignment
• Hidden Disagreement Detection
• Diagnostic Multiple-Choice Questions
• IoA-Suite
• IoA-Prober-8B