Overlap Number of Balls Model-Agnostic CounterFactuals (ONB-MACF): A Data-Morphology-based Counterfactual Generation Method for Trustworthy Artificial Intelligence

Counterfactual explanations help demystify AI decisions: The paper demonstrates that by showing users "what would need to change" for an AI system to make a different decision, we can make black-box AI systems more transparent and trustworthy—addressing a key concern in responsible AI deployment.

The ONB-MACF method generates more realistic alternative scenarios: Unlike some explanation methods, this approach ensures counterfactuals follow actual data patterns and require minimal changes to input features, making the explanations more practical and believable to stakeholders evaluating AI system decisions.

Model-agnostic approaches increase responsible AI accessibility: Because ONB-MACF works across different types of AI systems without needing access to internal model details, it democratizes explainability tools—allowing organizations to audit and understand AI decisions regardless of which specific algorithms they use.

Data structure matters for trustworthy explanations: The paper's focus on "data morphology" (how data is naturally distributed) shows that generating explanations grounded in real-world data patterns produces higher-quality insights than methods that ignore this structure, supporting more reliable AI auditing.

Comprehensive testing validates real-world applicability: The benchmark testing across multiple datasets and quality metrics provides evidence that this method works reliably in practice, which is essential for responsible AI practitioners deciding which explanation tools to implement in their systems.