An Ethical eValuation Agent (EeVA): Results of a Proof-of-Concept Test on a Prototype Agentic-like Workflow to Assist Ethical Deliberations

Researchers developed EeVA, an LLM-based workflow tool that assists non-specialists in conducting structured ethical deliberation across multiple frameworks rather than providing definitive answers. Proof-of-concept testing on three real-world cases demonstrated the system's ability to synthesize complex ethical perspectives, identify convergences and tensions, and communicate findings accessibly to non-ethicists.

EeVA addresses a significant gap in how organizations approach ethical decision-making when specialist expertise is unavailable. Rather than attempting to resolve moral disagreements—an impossible task—the system scaffolds comparative ethical reflection by evaluating scenarios against ten distinct frameworks simultaneously. The proof-of-concept results validate that LLMs can preserve ethical plurality while making sophisticated analysis accessible to non-specialists, a meaningful advancement in applied AI ethics.

This work emerges within broader recognition that ethical deliberation requires structured methodology rather than intuitive judgment. Most organizations lack ethics expertise on staff, yet face increasingly complex decisions around technology deployment, resource allocation, and stakeholder impacts. EeVA's three-workflow architecture demonstrates how modular AI systems can decompose ethics work into discoverable, auditable stages rather than treating it as a black-box recommendation engine.

The research has implications for AI governance, product development, and organizational decision-making. Companies building systems with ethical dimensions—particularly in emerging areas like autonomous mobility, energy systems, and social services—could adopt similar workflows to document deliberative processes and identify where cross-framework consensus breaks down. This transparency supports both internal stakeholder confidence and external accountability claims.

Future development requires human evaluation studies to verify whether non-specialists actually benefit from EeVA's outputs and whether the structured approach shifts organizational behavior toward more ethical outcomes. Reproducibility testing across diverse contexts and efficiency improvements for larger-scale deployment remain critical before mainstream adoption.

• →EeVA uses LLM workflows to scaffold structured ethical deliberation rather than deliver definitive moral answers, addressing accessibility gaps in ethics expertise.
• →Proof-of-concept testing across three real-world cases produced consistent, framework-specific evaluations that identified both convergences and persistent ethical tensions.
• →The system communicates complex multi-framework analysis in formats accessible to non-ethicists while preserving nuance and disagreement.
• →EeVA's architecture demonstrates how modular AI workflows can make ethics work auditable and transparent rather than opaque.
• →Further development requires human evaluation and user testing before the tool can mature beyond proof-of-concept stage.