Grounded Misunderstandings in Asymmetric Dialogue: A Perspectivist Annotation Scheme for MapTask
This provides a resource and analytic framework for studying grounded misunderstandings in collaborative dialogue, particularly for evaluating language models' ability to model perspective-dependent grounding, though it is incremental in applying existing methods to new annotation tasks.
The researchers tackled the problem of misunderstandings in asymmetric dialogue by developing a perspectivist annotation scheme for the HCRC MapTask corpus to track how understanding emerges and diverges, finding that full misunderstandings are rare after lexical unification but multiplicity discrepancies systematically cause divergences.
Collaborative dialogue relies on participants incrementally establishing common ground, yet in asymmetric settings they may believe they agree while referring to different entities. We introduce a perspectivist annotation scheme for the HCRC MapTask corpus (Anderson et al., 1991) that separately captures speaker and addressee grounded interpretations for each reference expression, enabling us to trace how understanding emerges, diverges, and repairs over time. Using a scheme-constrained LLM annotation pipeline, we obtain 13k annotated reference expressions with reliability estimates and analyze the resulting understanding states. The results show that full misunderstandings are rare once lexical variants are unified, but multiplicity discrepancies systematically induce divergences, revealing how apparent grounding can mask referential misalignment. Our framework provides both a resource and an analytic lens for studying grounded misunderstanding and for evaluating (V)LLMs' capacity to model perspective-dependent grounding in collaborative dialogue.