Can non-monotonic reasoning be efficient for practical applications? This paper explores that question by examining defeasible logic, presenting it as a computationally effective alternative to highly expressive but costly logics. The research introduces two new systems: a query-answering system using backward-chaining and a forward-chaining implementation that derives all possible conclusions. Showing the logic can handle large theories containing hundreds of thousands of rules, experimentation reveals defeasible logic has linear complexity, which dramatically contrasts with the exponential complexity of most other non-monotonic logics. The systems' efficiency and simplicity makes it an appealing modeling language for real-world applications. The study points to the potential of defeasible logic in modeling regulations and business rules, offering a pathway to computationally tractable non-monotonic reasoning.
Published in the International Journal on Artificial Intelligence Tools, this paper addresses the journal's focus on practical AI applications. By presenting efficient systems for defeasible logic, the research offers a valuable tool for AI practitioners. The work contributes to the journal’s ongoing exploration of AI methodologies.