Frontiers of Fairness: Socio-technical Challenges to Algorithmic Fairness and Justice

Artificial Intelligence (AI) systems are increasingly influencing critical decisions in domains such as healthcare, finance, law enforcement, and education and informing our understanding of medicine, science, and climate, amongst others. While these technologies promise transformative benefits, they also raise profound concerns about fairness and justice. Ensuring equitable outcomes in AI systems remains a formidable challenge, including but not limited to biases embedded in training data, complexities of translating human values into algorithms, multiple fairness definitions and metrics, and trade-offs between accuracy and fairness. Challenges to fairness and justice in AI systems are complex and nuanced: Technical approaches have been criticised as a panacea for ensuring fairness in their application. Yet, diverse cultural and societal contexts where perception of justice and fairness in process or outcomes vary, complexity in real-world scenarios in the scalability of fairness interventions and reckoning of a fairness to who benefits and who bears the risks of AI systems leave critical gaps in research, practice and governance to ensure AI systems are fair and just. This talk will highlight issues and open challenges in algorithmic fairness and justice, emphasising the need for interdisciplinary approaches, combining technical innovation with philosophical, sociological, and legal insights to build AI systems that serve humanity fairly and justly.