Encoding Organisational Judgement for Artificial Reasoners: Synthetic Reconstruction, Withholding, and Contrasting Engineering Practices

Engineering organisations rely on consequential knowledge that is never fully written down and is ordinarily acquired through participation. This reproducible technical note makes that missing layer experimentally manipulable. A preliminary narrative-reconstruction study develops a two-stage apparatus and finds sensitivity to correct guidance without discrimination between an organised representation and proposition-matched checklists. A principal withholding study gives two fictional organisations contrasting local engineering practices and demonstrates under-performance when local judgement is absent, recovery from social history, checklists and an organised representation, and predicted harm from the wrong practice's representation. The combined result supports accessibility and correctness while rejecting a distinctive model-input interpretation of stems in these apparatuses.