AI Safety and Fairness

The session will explore critical challenges in developing trustworthy AI systems, focusing on two interconnected domains: AI safety and fairness/bias mitigation. We’ll examine the fundamental problem of ensuring AI systems pursue intended objectives without causing unintended harm. The session will also address systemic bias in AI systems, exploring how training data, algorithmic design, and deployment contexts perpetuate social inequalities across domains like criminal justice, or healthcare.

Drawing on interdisciplinary research from computer science and social science, participants will analyse case studies, evaluate current technical approaches to these challenges, and critically assess the social and ethical implications of proposed solutions.

Outcomes

Participants will:

• Identify key concepts and challenges related to AI safety and algorithmic fairness in contemporary AI systems.
• Analyse case studies to uncover how bias and safety risks emerge through training data, model design, and deployment contexts.
• Evaluate current technical and policy approaches to mitigating safety concerns and systemic bias in AI.
• Critically assess the social and ethical implications of deploying AI systems in domains such as healthcare, criminal justice, and education.
• Reflect on the role of interdisciplinary perspectives (from computer science and social sciences) in shaping more trustworthy and equitable AI systems.

Contributors

Marco Ortolani is Senior Lecturer in Responsible AI, Keele University formerly assistant professor at the University of Palermo (Italy) and visiting researcher on a Fulbright grant at the Missouri University of Science and Technology. Dr Ortolani specialises in responsible and human-centred AI, with a strong interest in medical applications. His current research interests focus on developing robust methods for aligning advanced AI systems with human values, combining rigorous theoretical alignment, oversight and control techniques, to help ensure AI remains safe, reliable and beneficial in high-stakes domains

Baidaa Al-Bander is a Lecturer in AI and Data Ethics, Keele University. Dr Al-Bander has experience in building explainable and fair AI systems for healthcare. Baidaa Al-Bander’s work has advanced AI techniques for retinal image analysis, particularly in DR, diabetic macular edema and glaucoma using CFP and OCT image. Her current work focuses on ensuring AI systems are ethical, fair, transparent and usable in real-world healthcare settings.