# Reasoning Engine **Domain:** Artificial Intelligence / Architecture **Doc Type:** Canonical Concept Node **Maturity:** Developed ## Definition A **reasoning engine** is the computational machinery that applies inference procedures to represented facts, rules, hypotheses or models. ## Expert-System Context In classical expert systems, the engine is conceptually separated from the knowledge base. The engine determines how rules fire or hypotheses are searched; domain knowledge determines which distinctions and constraints exist. [[articles/The Evolutionary Roots of Silicon Valley|The Evolutionary Roots of Silicon Valley]] uses DENDRAL to show why the engine alone is insufficient. General search becomes useful only when the problem representation excludes impossibilities and preserves scientifically meaningful relations. ## Contemporary Context Modern learned models blur the boundary because representations and inference behavior are distributed through trained parameters. Tools, retrieval and system prompts reintroduce explicit layers. The governance question remains: which component made a consequential claim, and can its provenance be inspected? ## Key Insight **Reasoning power is inseparable from the representations, constraints and evidence through which the engine is allowed to reason.** ## Sources / Provenance - Edward Feigenbaum and Bruce Buchanan, DENDRAL and expert-systems literature. - Nexus synthesis: [[articles/The Evolutionary Roots of Silicon Valley|The Evolutionary Roots of Silicon Valley]]. ## See Also [[wiki/Encoded Domain Knowledge|Encoded Domain Knowledge]], [[wiki/Expert System|Expert System]], [[wiki/Inference|Inference]], [[wiki/Moral Accountability Layer|Moral Accountability Layer]]