TrustworthyAI Under Uncertainty: From Technical Performance to Human Reliance Zacharoula Papamitsiou

21/04/2026 This seminar explores what it means to build and use AI systems that are trustworthy under uncertainty. Moving beyond a narrow focus on accuracy, it examines how uncertainty, trust calibration, and human judgment shape appropriate reliance on AI in practice. The seminar will argue that trustworthy AI is a matter of designing and deploying systems that support well-calibrated human trust. We will focus on four contexts where AI is increasingly influential: education, media, business innovation, and defence applications. Across these domains, the key challenge is not only whether AI systems are accurate, but how humans and institutions interpret their limits, calibrate reliance, and make responsible decisions under uncertainty. The seminar will explore how AI can support judgment without obscuring risk, amplifying overconfidence, or undermining accountable human decision-making, and highlight the relational nature of trust.