From Policy to Practice: Research Directions for Trustworthy and Responsible Artificial Intelligence “By Design”

Rapid advancements in the development and adoption of artificial intelligence (AI) have accelerated the need for trustworthy and responsible AI (TRAI). National/international AI governance and risk management policies and frameworks have identified a core set of tenets for TRAI, including fairness, safety, privacy, security, transparency, explainability, and responsible deployment. Responsible AI processes/tools (RAPs) are solutions designed to operationalize and implement the tenets, serving as a middle layer between the tenets and real-world AI-embedded processes. In recent years, the design of RAPs has emerged as an important avenue for computational and social science researchers, practitioners, and policymakers. We highlight six important research directions for the design of RAPs. Using a real-world case study, we describe the importance of each research direction and illustrate current challenges.