Domain Project

We are pleased to announce that the paper “Greedy ABA Learning for Case-Based Reasoning” has been accepted for presentation at the 24th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2025) in Detroit, Michigan.

This work contributes to the growing area of logic-based learning by introducing a new variant of Assumption-Based Argumentation (ABA) learning tailored to case-based reasoning (CBR) settings. Traditional ABA learning methods involve highly non-deterministic search spaces, which pose challenges in scalability and interpretability. The proposed Greedy ABA Learning strategy addresses this by adopting a deterministic and exhaustive transformation sequence, effectively narrowing down the solution space while preserving explainability.

In coherent casebases—where identical features yield consistent labels—Greedy ABA Learning is proven to correspond exactly to AA-CBR, a well-known method for abstract argumentation-based case reasoning. More importantly, it also generalizes to handle incoherent casebases using credulous reasoning under stable semantics, enabling it to manage real-world data inconsistencies.

The paper presents formal results ensuring soundness, termination, and intensionality of the learning process, supported by illustrative examples. This research not only strengthens the theoretical foundations of symbolic learning from cases but also opens new avenues for scalable, interpretable reasoning in AI systems.

We look forward to presenting our results and engaging with the multi-agent systems community at AAMAS 2025.