Decisions and Public Policies

This lecture explores the use of machine learning tools in making decisions related to public policies, focusing on eligibility, rights, and actions. It discusses examples like COMPAS and SYRI, highlighting issues of fairness and accuracy. The controversy surrounding algorithms like SYRI in identifying social security fraud is also examined. Additionally, the lecture delves into the concept of profiling jobseekers to estimate long-term unemployment risks, showcasing tools like the Flemish model and the French 'Bob Emploi'. The evaluation of 'Bob Emploi' through a large randomized controlled trial is presented, along with discussions on discrimination, fairness, and the challenges of public acceptability in the context of public employment services.