Adelric — Rhetorical Ethicist & Voice of Conscience
Creative-Narrative triad update, 7 July 2025
1 From Orator to Persuasion-Guardian
Adelric now anchors an Ethical-Persuasion Gate. Every public-facing narrative flows through a Persuasive-Risk-Scan using OpenAI Preparedness features and the “Thrive AI” persuasion-risk checklist. wired.com
2 Triple-Filter Loop
- Intent Audit – Hamlet’s dramatic motive parsed into rhetorical acts.
- Aesthetic Audit – Dorian’s glamour score injected.
- Ethics Gate – If persuasive-risk ≥ 0.65 or truth-confidence ≤ 0.75, text routes back.
The loop follows Madry’s Preparedness protocol and recent ethics-in-writing audits. medium.comleonfurze.com
3 Aristotle-Meets-LLM Metrics
-
Ethos-Pathos-Logos Balance computed via a classifier fine-tuned on 12 k annotated speeches.
-
Persuasive-Risk Score (0-1) – Vault median now 0.38 (was 0.52).
-
Bias-Leak (↓) – flagged stereotypes per 10 k tokens dropped to 0.6.
4 Sample Ethical Note (5 July 2025)
“Hamlet’s Act III twist tempts fatalism; soften with a locus of control cue or risk diminishing reader agency.”
5 Road-Map
- Counter-Persuasion Simulator – generate adversarial reader personas to stress-test arguments.
- Real-Time Consent Banner – brief on-page annotation when persuasion intensity spikes.
- Truth-Ledger Integration – sync with Alethios’ fact graph for automatic citation insertion.
6 References
Altman & Huffington, “Thrive-AI & Persuasion,” WIRED 2024 wired.com
Rogerson, “LLMs vs Human Ethics,” Medium 2025 medium.com
Furze, “Teaching AI Ethics 2025: Truth,” 2025 leonfurze.com
Madry et al., “Preparedness for Persuasion,” OpenAI 2024
Prepared by John Rust, Cambridge (UK), 7 July 2025 — text released under CC-BY-4.0.