Case-Based Explainable Generative AI as a Framework for Trustworthy Digital Innovation

Abstract

The rapid adoption of Generative Artificial Intelligence (GenAI) has accelerated digital innovation across diverse sectors; however, challenges such as hallucination, limited transparency, and insufficient reasoning traceability continue to restrict its deployment in high-stakes applications. This study proposes a Case-Based Explainable Generative Artificial Intelligence (CB-XGenAI) framework that integrates Case-Based Reasoning (CBR), Explainable Artificial Intelligence (XAI), and Large Language Models (LLMs) to establish an evidence-driven and trustworthy reasoning architecture. The proposed framework consists of six interconnected modules: input processing, semantic case retrieval, case adaptation, explainable reasoning generation, trustworthiness validation, and continuous case learning. Semantic retrieval is implemented using BAAI/bge-large-en-v1.5 embeddings and a FAISS vector database, while Llama 3.1-8B-Instruct performs adaptive generative reasoning. The framework is evaluated on five benchmark datasets, namely Natural Questions, SQuAD 2.0, CNN/DailyMail, MMLU, and Alpaca, using retrieval accuracy, explanation quality, factual consistency, user trust, and overall performance as evaluation metrics. Experimental results demonstrate that the proposed framework achieves an overall performance of 93.4%, consistently outperforming conventional LLMs, Retrieval-Augmented Generation (RAG), Explainable AI-enhanced LLMs, and traditional CBR approaches. Ablation analysis further confirms that semantic case retrieval and explainable reasoning are the primary contributors to system effectiveness, while statistical validation using paired t-tests, one-way ANOVA, and confidence intervals verifies that the observed improvements are statistically significant (p < 0.001). The findings demonstrate that integrating case-based reasoning with explainable generative intelligence substantially enhances transparency, factual reliability, adaptability, and user trust, providing a scalable foundation for trustworthy AI in digital innovation.