XXAI: Towards eXplicitly eXplainable Artificial Intelligence

The "black box" problem remains a critical barrier to AI safety: The paper identifies that modern neural networks cannot explain their decisions explicitly, creating reliability and safety concerns that limit trustworthy AI deployment in high-stakes domains.

Symbolic AI offers transparency but faces practical limitations: While symbolic AI systems are inherently explainable ("white box"), they struggle with mathematical opacity, lack of unified knowledge frameworks, and computational complexity that has prevented their widespread adoption.

A new approach combines explainability with logical reasoning: XXAI proposes using deterministic logical cellular automata derived from first-principles domain knowledge, enabling AI systems to make decisions through transparent, step-by-step logical inference rather than opaque neural network calculations.

Explainable AI can verify neural network safety: The paper suggests XXAI could automatically check whether black-box neural network solutions meet reliability, security, and ethical standards during both training and deployment phases—addressing a key governance challenge.

Domain knowledge is foundational to trustworthy AI: By grounding AI decisions in explicit domain theories and general principles rather than pattern recognition alone, XXAI demonstrates that transparent, verifiable AI requires building knowledge systems from the ground up.