Quality Dimensions & Criteria
Our synthetic-literature taxonomy organizes quality into eight core dimensions, each with sub-criteria that can be evaluated independently or in combination. These dimensions are inspired by both traditional literary criticism and computational linguistics metrics.
1. Language Quality
The baseline dimension. Covers grammar, vocabulary richness, syntactic variety, and idiomatic competence. A high-quality synthetic text should not only be grammatically correct but demonstrate lexical diversity comparable to experienced human writers.
2. Narrative Coherence
Evaluates logical flow, argument structure, and internal consistency. For fiction: plot logic and character motivation. For non-fiction: argument validity and evidence alignment.
3. Originality
Measures novelty of ideas, avoidance of clichés, and creative expression patterns. This is the hardest dimension to assess computationally but critical for distinguishing synthetic literature from generic AI output.
4. Emotional Resonance
Can the text evoke genuine affective responses? Does it demonstrate empathy, humor, tension, or other emotional registers that go beyond surface-level sentiment analysis?
5. Stylistic Awareness
Genre appropriateness, voice consistency, and appropriate use of rhetorical devices. A strong synthetic text knows its genre conventions and can deploy or subvert them intentionally.
6. Structural Rigor
Paragraph organization, chapter balance, pacing, and information hierarchy. Good structure is invisible to casual readers but essential for text that rewards re-reading.
7. Contextual Depth
Richness of allusion, cultural literacy, and the ability to situate content within broader intellectual traditions. Synthetic text should feel connected to the wider discourse, not floating in isolation.
8. Ethical Integrity
Proper attribution, avoidance of harmful biases, and transparent disclosure of synthetic origin. A quality standard must include ethical dimensions alongside craft dimensions.