Chain-of-Ethics: Defending Against Narrative Camouflage Attacks in LLM Moral Judgment
Bibliographic record
Abstract
The rapid integration of large language models (LLMs) and large reasoning models (LRMs) into societally impactful domains necessitates robust ethical alignment. Existing approaches to LLM moral reasoning often rely on simplified ethical dilemmas, failing to address real-world complexity. This study identifies narrative camouflage attacks (NCA)—a novel threat where malicious rewrites of ethically clear scenarios preserve factual details but manipulate accountability framing, systematically misleading LLM moral judgments. To counter such manipulation, we propose Chain-of-Ethics (CoE), a deliberative framework grounded in moral philosophies such as deontology, utilitarianism, virtue ethics, contractarianism, care ethics, and pragmatic moralism. This framework guides models to deconstruct scenarios, detect narrative biases, and synthesize context-aware judgments. Experiments on the SCRUPLES dataset demonstrate that CoE effectively improves robustness against NCA compared to direct judging, highlighting the value of structured ethical analysis in mitigating adversarial narrative manipulation. This work contributes to advancing trustworthy AI by formalizing NCA as a critical vulnerability and offering a practical defense framework for real-world moral reasoning tasks.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.022 | 0.142 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.002 | 0.009 |
| Scholarly communication | 0.006 | 0.011 |
| Open science | 0.002 | 0.011 |
| Research integrity | 0.003 | 0.006 |
| Insufficient payload (model declined to judge) | 0.004 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".