Belief, Awareness, and Two-Dimensional Logic

Hu Liu, Shier JuIJCAI 2003

Abstract

Classical doxastic logic often suffers from the "logical omniscience" problem, assuming agents believe all logical consequences of their beliefs. This paper proposes a two-dimensional logic framework that incorporates awareness as a fundamental dimension alongside truth. By defining truth-degrees as pairs of truth and awareness values, the authors introduce Two-Dimensional Awareness Logic (TDAL) and its hybrid counterpart (HTAL), providing a robust mathematical foundation for modeling resource-bounded agents.

Introduction

In traditional epistemic and doxastic logics, the possible-worlds model defines belief as truth in all accessible worlds. A significant drawback of this approach is logical omniscience: if an agent believes $P$, and $P \implies Q$ is a tautology, the agent is modeled as believing $Q$. This fails to account for human or artificial agents with limited cognitive resources or conceptual gaps.

Previous solutions, such as Fagin and Halpern’s Logic of General Awareness (GAL), utilize an "awareness set" to filter beliefs. However, GAL is often criticized for being overly rigid and failing to distinguish between disbelief due to falsity and non-belief due to lack of conceptual awareness.

Two-Dimensional Logic Framework

The authors propose a shift from one-dimensional truth values to two-dimensional truth-degrees. Each formula $\phi$ is assigned a pair $(v, a)$, where $v \in \{1, 0\}$ represents the objective truth value and $a \in \{1, 0\}$ represents the agent's awareness of the concepts within the formula.

Logical Connectives and Contagion

In this framework, awareness is treated as "contagious." If an agent lacks awareness of any component of a complex proposition, the entire proposition may be treated as nonsensical or false from their perspective. This is formally captured in the truth matrices for the logical connectives.

Truth matrices for conjunction ($A \land B$) and implication ($A \to B$) in the 2D logic.

As shown in Figure 1, the conjunction of a known true statement (11) with a statement containing unknown concepts (10 or 00) results in a state of non-awareness (00). This skeptical approach mirrors how agents often dismiss information they cannot conceptually process.

Truth matrices for negation ($\neg A$) and disjunction ($A \lor B$) in the 2D logic.

Figure 2 further illustrates this behavior for negation and disjunction. Notably, even if one disjunct is true and known (11), the presence of an unknown disjunct (10 or 00) can render the entire disjunction "un-believable" (00) in the skeptical TDAL system.

Proposed Systems: TDAL and HTAL

The paper introduces two primary systems:

  1. Two-Dimensional Awareness Logic (TDAL): A cautious logic where awareness is strictly contagious across all connectives.
  2. Hybrid Two-Dimensional Awareness Logic (HTAL): A two-layered system. The "outer" layer follows classical logic for objective truth, while the "inner" layer (within the scope of the belief operator $B$) utilizes the 2D logic to represent the agent's subjective, bounded perspective.

Theoretical Results

The authors establish several key mathematical properties for these systems:

  • Soundness and Completeness: Axiom systems (System T for TDAL and System H for HTAL) are proven to be sound and complete with respect to their intended semantics.
  • Generalization of GAL: The research demonstrates that the classical Logic of General Awareness (GAL) is a specific, restricted case of the 2D framework, establishing the latter's superior generality.
  • Implicit Belief Preservation: While explicit belief is constrained to solve the omniscience problem, the logic preserves the standard properties of implicit belief, such as consistency and introspection.

Conclusion and Future Directions

This two-dimensional approach provides a flexible and mathematically rigorous way to model limited minds. By integrating awareness into the core logical mechanics rather than treating it as an external filter, it offers a more nuanced representation of subjective belief.

Future work includes extending the framework to multi-agent scenarios to model nested awareness (e.g., "I know that you don't know X"). Additionally, the computational complexity of reasoning within these 2D systems remains an open area for investigation, particularly for applications in automated theorem proving and AI.

Summary updated

BibTeX
@inproceedings{liu2003belief,
  author = {Hu Liu and Shier Ju},
  title = {{Belief, Awareness, and Two-Dimensional Logic}},
  booktitle = {International Joint Conference on Artificial Intelligence},
  year = {2003},
}
Figure