Illusion of Alignment: Detecting Hidden Disagreement in Collaborative Dialogue
This work addresses the critical problem of hidden disagreements in collaborative dialogue, which can lead to project failures, for anyone involved in collaborative tasks, including human teams and multi-agent systems. It is an incremental step in improving collaborative communication.
The paper identifies and quantifies the "illusion of alignment" (IoA) in collaborative dialogue, where participants appear to agree but harbor hidden disagreements. They developed a method using diagnostic multiple-choice questions to detect IoA, creating the IoA-Suite dataset and training IoA-Prober-8B, which achieved 51.8% F1 on IoA-Suite and surfaced 2.89 hidden disagreements per real meeting.
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.