How Case Based Reasoning Explained Neural Networks: An XAI Survey of Post-Hoc Explanation-by-Example in ANN-CBR Twins

Pairing neural networks with case-based reasoning systems offers a practical solution to AI explainability: The paper surveys over 1,100 studies showing that "ANN-CBR twins"—which combine neural networks with case-based reasoning—provide a coherent approach to explaining how AI systems make decisions, making this method applicable to both traditional and deep neural networks.

Explanation-by-example is more intuitive than abstract rules: Rather than generating complex mathematical explanations, this approach explains AI decisions by showing similar past cases the system considered, which aligns with how humans naturally justify their own reasoning and decisions.

The field lacks rigorous testing of whether these explanations actually help users: The paper identifies a critical gap in responsible AI research: while many explanation techniques exist, there has been insufficient "human user evaluation" to determine whether these explanations genuinely help people understand and trust AI systems in real-world settings.

Multiple explanation formats should be explored for different contexts: The research roadmap suggests testing various explanation approaches—including counterfactual examples (showing what would need to change for a different outcome), near-miss cases, and other formats—to determine which works best for different users and situations.

This approach bridges a fragmented research landscape: By systematically surveying scattered literature on explanation-by-example methods, the paper consolidates knowledge that can help standardize how we make AI systems more transparent and accountable.