An Epistemic Perspective on Agent Awareness
Pavel Naumov, Alexandra Pavlova
TL;DR
Addresses the problem of formalizing agent awareness as knowledge of existence and distinguishes de re and de dicto forms. Proposes two modalities $R$ and $D$ within a $2D$-semantics, integrating them with the standard $K$ modality. Develops a sound and complete axiomatization describing the interaction among $K$, $R$, and $D$, using a matrix-frame completeness proof and a canonical model. Introduces the notions of λ-assured sets and general awareness to manage epistemic constraints, with implications for reasoning about awareness in multi-agent systems. The results provide a rigorous semantic and proof-theoretic foundation for awareness in AI contexts, offering a precise framework for safety- and alignment-relevant reasoning about agent presence and properties.
Abstract
The paper proposes to treat agent awareness as a form of knowledge, breaking the tradition in the existing literature on awareness. It distinguishes the de re and de dicto forms of such knowledge. The work introduces two modalities capturing these forms and formally specifies their meaning using a version of 2D-semantics. The main technical result is a sound and complete logical system describing the interplay between the two proposed modalities and the standard "knowledge of the fact" modality.