Towards Explainable Agency in Multi-Agents Systems Using Inductive Logic Programming and Answer Set Programming

EasyChair Preprint 5465

Abstract

Logical reasoning is a fundamental aspect of human behaviour, and this is an important criteria to build human-like reasoning in intelligent autonomous multi-agents. So far, the field knowledge representation and reasoning have employed logic-based symbolic techniques to mimic the challenging task of incorporating human-like reasoning in multi-agent systems. However, the field of machine learning has shown increasing in- terest to take on this challenge. In this research, we describe a methodology which is based on Inductive Logic Programming and Answer Set Programming that enables autonomous agents to generate explanations and logic-based reasoning in a form hypothesis from a rich knowledge base (ontologies). Whilst this preliminary work addresses key limitations such as scalability and adaptability, we strongly emphasise the need for logic- based reasoning in multi-agents for interpretability and transparency in their behaviour.

Keyphrases: Answer Set Programming ·, Explainability, Inductive Logic Programming, Symbolic AI, logic-based reasoning, multi-agent systems, preference learning

@booklet{EasyChair:5465,
  author    = {Minal Suresh Patil and Kary Främling},
  title     = {Towards Explainable Agency in Multi-Agents Systems Using Inductive Logic Programming and Answer Set Programming},
  howpublished = {EasyChair Preprint 5465},
  year      = {EasyChair, 2021}}