LLM-Assisted vs Neuro-Symbolic Generation
Reference: This page is a source-based technical synthesis of the LaTeX chapter cited below. For exact syntax, availability or ABI signatures, verify the versioned source, manifest and executable tests.
Scope and source boundary
The two AI-assisted modeling paths differ mainly in how proposals are checked; neither eliminates semantic acceptance gates.
The supplied method distinguishes business specification, declarative conceptual modeling, semantic validation and technical operationalization. LLM assistance can propose names or structures, but it must not silently create validated domain authority. A neuro-symbolic route keeps candidate generation separate from model checking and acceptance.
Engineering rules
- Capture business vocabulary and observable scenarios before generating artifacts.
- Treat AI-generated content as candidates awaiting explicit validation.
- Apply type, role, identity, invariants and traceability checks.
- Project accepted models into Runtime contracts without discarding domain semantics.
Chapter outline (original LaTeX headings)
- Two Entry Points, Two Feedback Loops
- Path A: Model-First, H-Logic as a Semantic Synthesis
- Typical contexts for the model-first path
- Path B: H-Logic-First, Models as Operationalization
- Verbs and operationalization
- Typical contexts for the H-Logic-first path
LaTeX provenance
Primary chapter: Declarative Conceptual Programming with the logiCells Platform/developer-guide/chapters/two-paths-from-business-specifications.tex.