Verifier Scoring

This guide explains what Dramatica verifies, what it does not verify, and how to interpret verifier outputs during generation and revision.

What we verify

Dramatica verifies narrative intent alignment against the intended Storyform using:

What we do not verify

Dramatica does not directly score:

These are best handled by human judgment or separate graders.

How scoring works

Verifier outputs are split into two classes:

Constraints

Constraint checks act like compiler errors. A candidate can be rejected when it violates non-negotiable Storyform requirements.

Soft scores

Soft scores report degree of alignment across dimensions, such as:

These scores support ranking and revision, not style conformity.

How to interpret score movement

Interpretation should be relative within a controlled run:

Use score changes with qualitative inspection. Do not treat any single number as a full judgment of writing quality.

Common failure modes

Structurally valid but narratively flat

The draft may satisfy constraints while reading as emotionally inert. Keep verifier checks, then broaden candidate diversity and apply human taste filters.

Overfitting to structure

Drafts can become rigid when revision targets score movement without expressive variance. Preserve diverse candidates in selection.

Throughline leakage

Main Character and Influence Character pressures may blend in language even when structure seems intact. Treat this as an alignment failure and revise at the beat level.

Repetitive advancement

Storybeats can appear to advance numerically while repeating the same conflict expression. Add semantic progression checks in review.