EeVA: An LLM agent that assists ethical deliberation without giving answers

A team of researchers (Milford, Malgir, Vazquez) has introduced EeVA (Ethical eValuation Agent), an LLM-based agentic workflow designed to support ethical deliberation rather than provide definitive answers. Built in the visual programming tool n8n, EeVA uses three interconnected workflows—starter, worker, and emitter—to evaluate uploaded use cases against 10 distinct ethical frameworks. The system generates both framework-specific evaluations and integrated syntheses, shifting focus from simple right/wrong binaries toward design conditions, safeguards, and areas of irreducible tension.

In proof-of-concept testing on three published cases (urban mobility, peer-to-peer energy trading, and social-service resource allocation), EeVA produced consistently structured outputs that differentiated between frameworks and identified convergences and divergences. The syntheses were readable for non-specialists and recommended modifications to increase ethical alignment. The authors emphasize that EeVA is not meant to replace ethicists or resolve moral disagreements, but to scaffold structured deliberation—especially valuable where access to ethics expertise is limited. Future work is needed on reproducibility, human evaluation, and efficiency before it can be considered a mature tool.