Distributed Partial Information Puzzles: Examining Common Ground Construction Under Epistemic Asymmetry

Establishing common ground, a shared set of beliefs and mutually recognized facts, is fundamental to collaboration, yet remains a challenge for current AI systems, especially in multimodal, multiparty settings, where the collaborators bring different information to the table. We introduce the Distributed Partial Information Puzzle (DPIP), a collaborative construction task that elicits rich multimodal communication under epistemic asymmetry. We present a multimodal dataset of these interactions, annotated and temporally aligned across speech, gesture, and action modalities to support reasoning over propositional content and belief dynamics. We then evaluate two paradigms for modeling common ground: (1) an axiomatic pipeline based on Dynamic Epistemic Logic (DEL) that incrementally infers shared beliefs from multimodal updates, and (2) state-of-the-art large language models (LLMs) prompted to perform the same reasoning tasks. Results on the annotated DPIP data indicate that it poses a challenge to modern LLMs’ abilities to track both task progression and belief state. All data, annotations, and code will be publicly released to advance research on common ground tracking and multimodal epistemic reasoning.

Yifan Zhu, Mariah Bradford, Kenneth Lai, Timothy Obiso, Videep Venkatesha, James Pustejovsky and Nikhil Krishnaswamy . 2026. Distributed Partial Information Puzzles: Examining Common Ground Construction Under Epistemic Asymmetry. (to appear)