Two Sides of the Same Coin: Belief Revision and Enforcing Arguments

Adrian Haret, Johannes Peter Wallner, Stefan WoltranIJCAI 2018 · DOI

TL;DR

This paper introduces a formal, axiomatic framework for "enforcing arguments" within argumentation systems, treating it as a fundamental belief change operation. By establishing a duality between belief revision and argument enforcement, the authors provide a principled way to understand and implement argument enforcement, particularly in abstract argumentation frameworks.

What Problem Does This Paper Address?

Formal argumentation systems are designed to model debates and reasoning, where arguments are accepted or rejected based on their relationships (like attacks). A key operation in these systems is "enforcing arguments," which aims to ensure that certain arguments are accepted. This is typically done by making minimal changes to the argumentation system. However, a principled, axiomatic treatment of this operation, akin to the well-established axiomatic approach in belief change, has been lacking. This paper bridges this gap by formalizing argument enforcement as a belief change operator and exploring its deep connections to belief revision.

Methods

The authors tackle this problem by first axiomatizing propositional enforcement. They then establish a formal duality between belief revision and enforcement, showing that enforcement can be seen as the "dual" of revision. This duality is underpinned by novel representation results using rankings on sets of interpretations, which are more intuitive for enforcement than traditional rankings on individual interpretations. The paper then extends this axiomatic framework to abstract argumentation frameworks (AFs), adapting the axioms to account for the specific properties of AF semantics, particularly the concept of "adm-realizability" (whether a set of extensions can be realized by an AF).

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Enforcement as a Belief Change Operator: The paper successfully frames "enforcing arguments" not just as a mechanism within argumentation systems, but as a core belief change operation, opening it up to rigorous formal analysis.
  • Duality with Belief Revision: The most significant insight is the formal demonstration of a duality between belief revision and enforcement. This provides a powerful conceptual lens: while revision aims to incorporate new information by potentially discarding old beliefs, enforcement aims to ensure acceptance of certain arguments by finding a middle ground, preserving as much of the original state as possible.
  • Rankings on Sets of Interpretations: The introduction of rankings on sets of interpretations (rather than just individual interpretations) is a novel and effective way to capture the semantics of enforcement, offering a more intuitive representation than existing methods.
  • Adapting Axioms for AFs: The paper highlights the challenges and necessity of adapting propositional logic axioms to the specific constraints of argumentation frameworks (AFs), particularly the adm-realizability of extension sets. This demonstrates a nuanced understanding of how abstract logical principles must be tailored to concrete computational models.

Key Results

  • Axiomatization of Propositional Enforcement: The paper provides a set of axioms (E1-E6) for propositional enforcement and proves a representation result based on "e-assignments" (rankings on sets of interpretations).
  • Duality Theorem: Theorem 3 formally establishes the duality between propositional enforcement (▷) and revision (•) as: κ ▷ μ = ¬(«κ ο ¬μ).
  • Adaptation to Abstract Argumentation: The authors propose modified axioms for enforcement in AFs (E1-E6) and show how to obtain adm-realizable enforcement operators by refining the output of propositional enforcement operators using an "adm-closure" mechanism.
  • Concrete Enforcement Operators: The paper discusses how existing operators like drastic enforcement (▷D) and Hamming distance-based enforcement (▷H) can be adapted from revision to enforcement.

Limitations and Future Work

The authors acknowledge that adapting enforcement axioms to specific structural constraints on AFs (like "normal expansions") requires careful consideration of semantical characterizations, as direct application of propositional axioms may fail. Future work suggested includes extending the results to other AF semantics beyond admissible semantics and exploring further characterizations of structural changes in AFs. The paper also hints at the need to consider how enforcement operators behave when faced with agents who have beliefs and an argumentation framework, and how to restore consistency in such scenarios.

Summary updated

BibTeX
@inproceedings{haret2018two,
  author = {Adrian Haret and Johannes Peter Wallner and Stefan Woltran},
  title = {{Two Sides of the Same Coin: Belief Revision and Enforcing Arguments}},
  booktitle = {International Joint Conference on Artificial Intelligence},
  year = {2018},
  doi = {10.24963/ijcai.2018/256},
}
Figure