Hampton Court Maze Pilot

Abstract The Hampton Court Maze Pilot provides a minimal, operational demonstration of the three‑layer architecture for human‑interpretable machine reasoning. Using a simplified digital model of the Hampton Court Maze, the system separates problem solving, trace reconstruction, and human‑level interpretation into independent modules. A solver navigates the maze without explanation features, a trace system records its full decision path including dead ends and pivot points, and an interpreter converts the trace into a coherent narrative that a human reviewer can map onto the physical maze. This pilot validates the architecture’s core claim: machine reasoning can be decomposed into transparent, auditable layers that preserve rejected alternatives and support accountable human oversight.