Responsible Artificial Intelligence (RAI) is about ensuring that socioenvironmental responsibility is a fundamental and permeating consideration in all stages (conception, evaluation, deployment, monitoring, etc.) of AI development and governance. This is necessary to address and prevent harms and injustices, setting the course of AI development in a direction of sustainable benefit to our interconnected world. RAI requires a thoughtful and pragmatic synthesis of approaches from wide-ranging fields. The goal of this course is to equip students with the qualitative, quantitative, critical, reflective, and practical tools to bridge the gap between theory and practice, in order to make responsible AI a reality.

The course covers an evolving set of relevant topics such as user studies, participatory design, statistical significance testing, model comparison, generalization, randomized control trials, bias, interpretability and explainability, FAccT (fairness, accountability, transparency), equity, ethics, safety, alignment, robustness, scientific communication, stakeholder consultations, data governance, power and exploitation, digital labour, peer review, reliability engineering, information security, privacy, verification, auditing, reproducibility, red-teaming, unit-testing, sandboxing, scenario planning, risk, and impact analysis.