Trustworthy and Socially Responsible AI

Trustworthy and Socially Responsible AI studies how to design, analyze, and deploy AI systems that are verifiably reliable, fair, transparent, and abide by relevant human norms. Research in this area examines foundational questions such as robustness to distributional shifts, incomplete/inconsistent/noisy information, and adversarial inputs; verification of behavioral safety; interpretability of learned models; fairness and bias mitigation; privacy preservation; and accountability in automated decision-making systems. A key challenge is to ensure that AI systems behave predictably and safely in complex, multi-agent, high-stakes, and socially embedded environments, even when data are limited, dirty, compromised, or evolving over time.