Machine learning (a subfield of artificial intelligence) is now routinely used by policymakers in the design, calibration and evaluation of public policies. Solving these prediction policy problems requires tools that are tuned to minimise prediction errors, but also frameworks to ensure that models are fair.
Thanks to funding from the Dutch government, three researchers at UNU-MERIT (Stephan Dietrich, Michelle Gonzalez Amador and Alex Hunns) have developed Machine Learning for Public Policy (ML4PP), a free online course designed for university students that will start on Thursday 30 November 2023. In six sessions spread over three months, participants will be led through basic theory and applications of machine learning to public policy analysis in an interactive, self-paced course that culminates in a group-based 'Collaborative Policy Challenge' set by an expert guest speaker.
Participants in the course will:
- Gain a basic understanding of supervised machine learning algorithms
- Learn about the benefits and risks of applying these models to public policy issues
- Develop the necessary skills to train and assess the performance of machine learning algorithms for solving public policy problems
- Collaborate in international and multi-disciplinary teams
Students will have access to short video lectures from experts in the field, suggested reading, and short self-assessment assignments. Topics covered will include an introduction to the programming languages R and Python, linear classification and selection models, shrinkage models and cross-validation, tree-based models and fairness in prediction models.
The course will end with a Collaborative Policy Challenge: during a kick-off event with an expert practitioner from the field, interdisciplinary groups of students will be given a policy challenge to solve over the course of a fortnight, using the concepts and methods covered in the online course. At the final meeting, the teams will provide their possible solutions to the challenge and receive feedback from peers and policy experts.
Machine Learning for Public Policy is open to students of all backgrounds with a willingness to engage with both the topics and coding languages in the course.
Photo by James Harrison on Unsplash
