New arXiv Paper Proposes Taxonomy for Misunderstanding Detection in AI-Mediated Communication

A new paper on arXiv (2608.13604) addresses the urgent problem of misunderstanding detection in communication, which has increasingly shifted to AI-mediated channels. The authors analyze misunderstanding as a layered process where divergence is generated, amplified, and either detected and repaired or left unnoticed. They consolidate accounts from nine fields of research that typically do not cite each other, identifying eleven exact failure modes. Each failure mode operates at a specific point in the communicative process, leading to eight analytical layers derived from the literature. Eight mechanisms primarily generate divergence, two amplify existing divergence, and one is not specified in the abstract. The paper aims to provide a cross-disciplinary taxonomy and modeling framework, from pragmatics to AI agents, to improve detection and repair of misunderstandings in digital communication.