Artificial Intelligence / AI 0240 · Capstone · 2–3 minutes
The Succulent Audit
to the StudyWalks catalog
One full case — derivation, proof, and post-mortem — shows exactly where a rule system's confidence comes from and exactly where it cannot be trusted.
An autumn succulent arrives at the clinic: leaves yellowing on schedule, soil soggy from an enthusiastic new owner, decorative pot without drainage. Run bottom-up: rule one fires — overwatered; rule two fires — root rot risk; rule four fires — needs repotting. Fixed point. Run top-down on the repotting question and the proof tree grows cleanly to facts at every leaf: the system will answer why with a flawless trace. Now the post-mortem, because the diagnosis is wrong — this species yellows every autumn, watered or not. Walk the audit. The facts? Correctly logged: the leaves are yellow, the soil is soggy, the pot has no holes. The procedure? Sound and complete — the guarantee held; every derived atom is a true consequence of the knowledge base. The rules? There is the fault: rule one is honest botany for most plants and false for this one, and the machinery has no way to know which plant it is holding. Read the verdict carefully: the proof tree is real, the audit trail is real, the confidence is earned — with respect to the rules as written. Soundness transmits truth; it cannot mint it. When a rule system is wrong, it is wrong the way this succulent was misdiagnosed: upstream, in the written knowledge, with every downstream step in perfect order. That is why auditing such a system means auditing its rules — and why the next unit teaches rules that can hedge.
The trace's flawlessness is what makes the failure dangerous — confidence and correctness parted ways at the knowledge, not the logic.