When Repair is not Enough: Systematic Mitigation of Incomplete Knowledge Graphs

Safety analysis of knowledge graphs (KGs) extracted from system documentation is challenging because documentation is often incomplete or ambiguous. Existing validation under open-world reasoning misses hazards implied by missing information and does not provide a systematic path from hazard identification to minimal, traceable mitigation. We use Kripke structures to formalize KG semantics by enumerating feasible states and classifying outcomes into four cases: satisfaction, repair, mitigation, and unrealizability. For repair and mitigation, we combine reachability analysis with delta debugging to compute minimal predicate-level changes and synthesize a weakest safe completion that excludes unsafe terminal states while preserving maximal safe design freedom. The paper contributes a decision taxonomy for post-identification action, an optimization-based synthesis of additions and flips for minimal mitigation, and an unrealizability procedure that guides controlled safety-property relaxation with explicit traceability to documentation and safety artifacts. For unrealizable specifications, the procedure explicitly escalates from model-level correction to stakeholder-level action, assigning follow-up decisions to the appropriate actors (e.g., safety engineer, system designer, and user/operator) through safety-case traceability. In an automotive seatbelt case study, the framework identifies latent hazards, synthesizes minimal mitigation actions, and separates system-actuated from occupant-actuated causes, thereby improving the practical value of the analysis for assurance and certification.

Keywords: Safety-critical systems, knowledge graph mitigation, specification-level repair, delta debugging.