Identifying Roles, Requirements and Responsibilities in Trustworthy AI Systems

Domain practitioners need better visibility into AI systems they're deploying. The paper identifies a critical gap where experts in fields like healthcare and education are using complex AI systems without understanding the underlying data or how these systems work, highlighting the need for improved transparency mechanisms.

Multiple stakeholders have distinct responsibilities in trustworthy AI. The research identifies and maps out different roles involved in AI system deployment, each with specific requirements and responsibilities, suggesting that trustworthy AI requires clear accountability structures across the entire system lifecycle.

There's a fundamental tension between transparency and privacy that must be resolved. The paper highlights that providing domain practitioners with sufficient insight to assess AI systems appropriately often conflicts with privacy protections, indicating this is a key challenge that responsible AI deployment must actively address rather than ignore.

Domain expertise is essential for responsible AI deployment decisions. By centering the perspective of practitioners in critical fields, the paper emphasizes that technical AI developers alone cannot determine whether systems are appropriate for real-world use—domain experts must have the tools and information to make informed deployment decisions.