From data model to test evidence: Human-in-the-loop model-driven test generation for GxP-grade systems

Presented by: Hans Christian de Raad, OpenNovations

This poster presents a model-centric approach to automated testing for high-integrity systems in which the information model, business rules, workflow states, authorization policies, and human-defined invariants drive the generation of both application artifacts and multi-layer test evidence. Rather than treating testing as a separate activity added after implementation, the approach generates persistence-layer objects, REST API behavior, UI components, and systematic good, fuzzy, and false test families as part of the same controlled change process.

GxP-grade environments provide the stress case for this work, but the underlying pattern is broader: any domain facing sustained change pressure together with strong demands for traceability, security, data integrity, and controlled release can benefit. The poster shows how this model-driven test generation approach, with clear MBT relevance, supports cross-layer verification of workflows, permissions, integrity rules, and rejection behavior while keeping release judgment under human control.

It also highlights how machine logic already contributes through rule-based test derivation, structured boundary exploration, CI/CD execution, and evidence-by-construction, with AI positioned as a future augmentation for anomaly triage and boundary-case suggestion. Workflow diagrams and technology-stack references connect the architecture to practical tooling across PHP, Python, PostgreSQL, and browser-based testing.