Introduction to Fairness, Accountability, and Transparency in Machine Learning
MSc Option course, Hilary Term
- Reading list: Introduction to Fairness, Accountability, and Transparency in Machine Learning reading list
The purpose of this advanced machine learning course is to equip you with the technical and conceptual tools to be able to engage deeply and critically with machine learning research and practice. Although the machine learning methods covered are advanced, the course will also be structured to accommodate participants with minimal programming and mathematics experience.
The course strives to integrate historical and cultural context with contemporary methods in machine learning and scholarship on Fairness, Accountability, and Transparency (FAT*). The basic strategy is to ground the course in issues of public concern related to machine learning and build practical knowledge through engagement with the technical workings of algorithms related to those concerns. The lectures will primarily provide context, while the formative assessments provides opportunities to scaffold technical understanding and critically reflect.
The particular methods covered in the course have been selected in part to be complementary to what is covered in typical introductory machine learning courses, in part because of their importance in contemporary machine learning practice and research, and in part because of their relevance to major social issues.
The core goal of each class will be to ask, in addition to the mathematical or statistical assumptions being made by a given machine learning model, what are the social and psychological assumptions associated with the use of that model in particular situated applications? And what are the varied consequences of those assumptions in those particular social contexts?
The course will be taught from within a framework of anti-racism, intersectional feminism, and anti-oppression.
- Historical context of modern machine learning
- Race and machine learning
- Gender and machine learning
- Technical approaches to fairness
- Radical reimaginings of machine learning
- Tools for examining and communicating
- Organizing work and social movements
At the end of this course students will:
- have a technical understanding of sources of bias and discrimination in machine learning.
- have a conceptual understanding of how issues such as bias and discrimination in machine learning are linked to broader structures of power in society.
- know how to evaluate interventions for improving fairness, accountability, and transparency.
- be equipped with tools for imagining new, radical interventions in machine learning.