ExplainabilityTransparencyTrustworthinesscs.AI
Machine Reasoning Explainability
Kristijonas Cyras, Ramamurthy Badrinath, Swarup Kumar Mohalik, Anusha Mujumdar, Alexandros Nikou, Alessandro Previti, Vaishnavi Sundararajan, Aneta Vulgarakis Feljan
September 1, 2020
arXiv.org
16 citations
1 influential
Key Takeaways
-
Symbolic reasoning offers transparency advantages: Machine Reasoning uses symbolic approaches to formalize abstract reasoning, which inherently provides clearer explanations compared to black-box AI systems, making it particularly valuable for applications where understanding how decisions are made is critical.
-
Explainability has been a longstanding concern in AI: The paper notes that early Machine Reasoning research already prioritized Explainable AI (XAI), demonstrating that transparency concerns predate current AI debates and that decades of research exist on this challenge that modern AI systems can learn from.
-
Multiple AI subfields benefit from MR explainability techniques: Argumentation, constraint programming, logic programming, and planning systems all leverage Machine Reasoning explainability methods, suggesting that insights from symbolic AI could be broadly applicable across different types of AI systems.
-
Current XAI landscape may be missing established solutions: By providing an overview of MR explainability research, the paper suggests that the contemporary AI community's approach to explainability could be complemented or improved by incorporating techniques and lessons from decades of Machine Reasoning work that may have been overlooked.
Abstract
As a field of AI, Machine Reasoning (MR) uses largely symbolic means to formalize and emulate abstract reasoning. Studies in early MR have notably started inquiries into Explainable AI (XAI) -- arguably one of the biggest concerns today for the AI community. Work on explainable MR as well as on MR approaches to explainability in other areas of AI has continued ever since. It is especially potent in modern MR branches, such as argumentation, constraint and logic programming, planning. We hereby aim to provide a selective overview of MR explainability techniques and studies in hopes that insights from this long track of research will complement well the current XAI landscape. This document reports our work in-progress on MR explainability.