A Logic Programming Framework for Possibilistic Argumentation with Vague Knowledge

Carlos Iván Chesñevar, Guillermo Ricardo Simari, Teresa Alsinet, Lluís GodoUAI 2004

TL;DR

This paper introduces P-DeLP, a novel logic programming language that extends Defeasible Logic Programming (DeLP) to handle both vague knowledge (using fuzzy logic) and possibilistic uncertainty. It provides a formal framework for qualitative reasoning in the presence of incomplete, contradictory, and uncertain information, crucial for commonsense reasoning applications.

What Problem Does This Paper Address?

Traditional logic programming and even existing defeasible argumentation frameworks like DeLP struggle to represent and reason with imprecise or uncertain information. While DeLP handles defeasibility (exceptions to rules), it doesn't explicitly model uncertainty or vagueness in the knowledge itself. This limits its applicability in real-world scenarios where information is often fuzzy (e.g., "engine speed is low") or uncertain (e.g., "it's likely that the engine is okay"). This paper aims to bridge this gap by integrating fuzzy logic and possibilistic uncertainty into a defeasible logic programming framework, enabling more robust qualitative reasoning.

Methods

The core of P-DeLP is built upon Possibilistic Gödel Logic (PGL), which combines Gödel fuzzy logic for handling vagueness with a possibilistic framework for uncertainty.

Key methodological contributions include:

  • Fuzzy Propositional Variables: Introducing fuzzy propositional variables to represent vague statements (e.g., engine_speed(low)).
  • Possibilistic Uncertainty: Assigning necessity measures (values between 0 and 1) to formulas to quantify the certainty of propositions.
  • P-DeLP Language: Extending the DeLP language with fuzzy atoms and weighted clauses.
  • Argumentation Framework: Developing an argument-based inference engine. Arguments are tentative proofs for a conclusion, supported by a set of uncertain, weighted formulas.
  • Dialectical Trees: Employing dialectical trees to formalize the process of warrant computation. This involves identifying counterarguments, defining defeat conditions based on necessity measures, and recursively analyzing argument exchanges to determine which arguments are ultimately warranted (undefeated).
  • Proof Method: Utilizing a generalized modus ponens rule from PGL for deduction, ensuring completeness for determining the maximum degree of possibilistic entailment.

Datasets

Not discussed in the paper. The paper focuses on formalizing a framework and provides an example (Peng) to illustrate its concepts, rather than empirical evaluation on specific datasets.

Key Insights and Ideas

  • Bridging Vagueness and Uncertainty: The paper successfully integrates two distinct forms of imprecise information (fuzziness and possibilistic uncertainty) within a single logical framework, which is a significant step towards more human-like reasoning.
  • Argument Strength as Necessity Measure: Using necessity measures as the primary criterion for comparing and resolving conflicts between arguments provides a principled way to handle uncertainty in defeasible reasoning.
  • Dialectical Analysis for Warrant: The extension of dialectical trees to handle weighted arguments offers a robust mechanism for determining justified conclusions in the face of conflicting, uncertain information.
  • Flexibility of PGL: The choice of PGL as the underlying logic is justified by its compatibility with fuzzy events and its ability to support efficient proof methods, making it suitable for logic programming extensions.

Key Results

The paper presents the formal definitions for the P-DeLP language, its possibilistic model, and the inference mechanism for computing warranted arguments. The key result is the formalization of a framework that allows for qualitative reasoning with both vague and uncertain knowledge, extending the capabilities of existing defeasible logic programming systems. The paper demonstrates this through an example (Peng) illustrating how arguments are constructed, how conflicts are identified, and how warrant is computed.

Limitations and Future Work

  • Limitations:
    • The current framework is propositional. The authors mention the possibility of extending it to first-order logic, which would allow for fuzzy unification and more expressive knowledge representation.
  • The paper focuses on the theoretical framework. While it mentions the successful integration of DeLP in real-world applications, it doesn't present an implementation or empirical evaluation of P-DeLP itself.
  • Future Work:
    • Implementation: Extending the existing DeLP implementation to incorporate P-DeLP's features.
  • Applications: Applying P-DeLP to improve existing real-world applications and develop new ones.
  • First-Order Extension: Extending P-DeLP to first-order logic, incorporating fuzzy unification.
  • Aggregated Preference Criteria: Exploring the combination of necessity measures with other syntax-based criteria (like specificity) for argument comparison.

Summary updated

BibTeX
@inproceedings{chesnevar2004logic,
  author = {Carlos Iván Chesñevar and Guillermo Ricardo Simari and Teresa Alsinet and Lluís Godo},
  title = {{A Logic Programming Framework for Possibilistic Argumentation with Vague Knowledge}},
  booktitle = {Conference on Uncertainty in Artificial Intelligence},
  year = {2004},
}
Figure