The adoption of artificial intelligence (AI) technologies in high-stakes fields like healthcare, finance, and aviation requires accurate and reliable systems that provide traceable and interpretable results. This chapter focuses on the crucial features of knowledge representation (KR) and explainable AI (XAI), calling for hybrid systems that integrate both sub-symbolic and symbolic reasoning. Traditional models of reasoning symbolically provide high interpretability but lack scalability, while modern models based on neural computing provide high performance but lack explainability. We review the literature on neuro-symbolic systems, probabilistic graphical models, attention mechanisms in neural networks, and embeddings of knowledge graphs trained on multi-layer perceptrons. We propose an adapted multi-layer Gradient-weighted Class Activation Mapping using EfficientNet-B3 for interpretable visual explanation in land-use scene classification. Key themes include mutual information in online advertising, large language models’ transparency, and interactive explanation systems. We classify XAI techniques based on their granularity, integration timing, and functional type, and discuss the contribution of KR learning to AI interpretability. Emphasizing human- and machine-centric explanations, this work promotes responsible, ethical AI design that aligns with cognitive expectations and regulatory standards.