Jury Consensus Model
How it works
The Jury evaluation process:
- Your text is submitted to multiple AI models simultaneously
- Each model analyzes the writing independently using the same craft dimensions as the Diagnostic
- Individual assessments are collected
- A consensus layer synthesizes the results, identifying agreement and divergence
- The final output shows consensus scores, per-model perspectives, and areas of disagreement
Why multiple models
Writing quality is subjective. A passage one editor finds tightly paced, another might find rushed. A metaphor one reader finds fresh, another might find clichéd. Single-model analysis gives you one perspective; multi-model analysis gives you several.
When multiple models independently identify the same issue — "the transition at paragraph 3 is abrupt" — that's high-confidence feedback. When models disagree — one says the opening is strong, another says it's slow — that tells you the opening is ambiguous or polarizing, which is itself valuable information.
Consensus scoring
For each craft dimension, the Jury reports:
- Consensus score — the average assessment across all models
- Agreement level — how closely models aligned (high = uniform assessment, low = divided opinions)
- Notable divergences — specific cases where one model's assessment differed significantly from the others
High agreement on a weakness is a clear signal to revise. Low agreement may mean the passage is genuinely ambiguous — different readers will react differently, and you should decide whether that ambiguity is intentional.
When to use the Jury
The Jury is most valuable on polished work — writing you've already revised and want to evaluate rigorously before considering it complete. It's more expensive and slower than the Diagnostic (multiple model calls), so it's best used selectively rather than on every draft.
Recommended workflow: iterate with Diagnostic during drafting → submit to Jury when you believe the piece is near-final → use Jury feedback for the final revision pass.