Dr. Shion Guha, Co-Director of GRACE and faculty at the University of Toronto, will deliver a keynote exploring how predictive risk models reinforce inequality, and why AI cannot alone resolve complex social challenges.

On 25 September 2025, the London Public Library officially launched The Glass Room, an interactive exhibit designed to help community members engage with the technologies woven into daily life. The Glass Room invites visitors to question what data is collected about them, how trustworthy information is determined, and how power is allocated in algorithmic systems. The exhibit will be open to the public for four months.

To mark the opening, Dr. Guha delivered “AI Won’t Fix Our Social Problems: Deconstructing Risk in Predictive Risk Models.” In his talk, he argued that many predictive risk models now used in sectors like child welfare, criminal justice, homelessness services, and immigration do not reliably forecast harm. Instead, they often reproduce the patterns of institutional decision-making—turning agency actions into proxies for risk. He traced how labeling choices, data sources, and integration into workflows create bias, conceal error, and shift burdens onto communities already under surveillance.

Dr. Guha also shared a human-centered data science framework developed through his work. This approach integrates statistical methods with interpretive inquiry, participatory auditing with affected communities, and governance checkpoints. Rather than scoring speculative risk, the framework emphasizes designing systems toward clear safety and support objectives. Although Dr. Guha grounded much of his analysis in child welfare, he stressed that the concepts apply broadly to other public services where administrative actions masquerade as objective truth.

For full details, including the event abstract, speaker biography, schedule, and location, read more on the Starling Centre website.