Modeling Ambiguity in a Multi-Agent System
Christof Monz
Abstract
This paper investigates the formal pragmatics of ambiguous expressions by modeling ambiguity in a multi-agent system. Such a framework allows us to give a more refined notion of the kind of information that is conveyed by ambiguous expressions. We analyze how ambiguity affects the knowledge of the dialog participants and, especially, what they know about each other after an ambiguous sentence has been uttered. The agents communicate with each other by means of a TELL-function, whose application is constrained by an implementation of some of Grice's maxims. The information states of the multi-agent system itself are represented as a Kripke structures and TELL is an update function on those structures. This framework enables us to distinguish between the information conveyed by ambiguous sentences vs. the information conveyed by disjunctions, and between semantic ambiguity vs. perceived ambiguity.
Create a lesson
Related papers
English Calibration Erodes Lithuanian in Expert-Pruned Mixture-of-Experts Language Models
Tomas Rasymas, Povilas Kvedaras
The Gating Mechanism of KV Cache Eviction
Hao Gu, Xintong Yang, Zhiwei Li et al.
Fine-Tuning a Transducer for Impaired Speech: Where the Gains Appear and What They Mean for Evaluation
Kien Nguyen, Tuan Dinh, Bao Phan
Stress Testing Chain-of-Thought Monitoring Against Covert Misalignment
Yao Huang, Yifan Wang, Yitong Sun et al.
SKIM: Pruning Large Language Model Agents via Selective Knowledge Informed Masking
Moonseok Choi, Jongwon Jeong, Minki Kang et al.
The Memory Trap: Language Agents Overgeneralize from Benign Interactions
Kaiqu Liang, Jaime Fernández Fisac, Thomas L. Griffiths et al.