E-LENS: User Requirements-Oriented AI Ethics Assurance

Ethics assurance requires structured frameworks similar to aviation safety: The paper argues that AI ethics needs formal assurance approaches comparable to those used in safety-critical industries like aviation, rather than relying solely on ethical principles. This structured approach helps move ethics from abstract guidelines to concrete, verifiable practices.

Three pillars—user requirements, evidence, and validation—are essential: The E-LENS framework centers on understanding what stakeholders actually need from AI systems, gathering concrete evidence that ethical standards are met, and systematically validating compliance. This ensures ethics assurance is grounded in real-world user concerns rather than theoretical ideals.

Current AI ethics frameworks fall short in real-world applications: The paper identifies a significant gap between the proliferation of AI ethical principles and the ability to actually assure ethics in deployed systems, suggesting that existing frameworks lack practical implementation mechanisms for organizations building AI products.

Hazard analysis methods from safety engineering can be adapted for ethics: By borrowing proven techniques from aviation and other safety-critical domains, the approach helps identify potential ethical risks in AI systems before deployment—demonstrated through a case study of an AI-driven hiring system.

Practical implementation requires dedicated platforms and tools: The E-LENS platform demonstrates that ethics assurance needs technical infrastructure to document requirements, collect evidence, and validate compliance, making responsible AI development a systematic process rather than an ad-hoc consideration.