Skip down to main content

Fairness, Accountability, and Transparency in Machine Learning

Key Information

Course details
MSc Option course, Hilary Term
Reading list
View now
Dr Luc Rocher, Dr Adam Mahdi



Machine Learning (ML) and Data Science are increasingly used to mine human data, derive insights, automate decision making in social settings, or build new digital representations of our behaviours. This course aims to critically analyse how ML technologies can both bring social progress but also build and reinforce racism, discrimination, and injustice. We will draw from an interdisciplinary literature at the intersection of computer science and social science.

The purpose of this course is to equip students with the technical and conceptual tools to be able to engage deeply and critically with machine learning as a phenomenon. The course will ground, motivate, and contextualize these issues in the experiences of individuals and communities impacted most by automated decision making, data mining, and digital technologies. The students are expected to engage both through the critical lens of their own experience and through the materials discussed in class.


  • Knowledge, discrimination, and power by humans and machines
  • Measuring and addressing social injustice in models and algorithms
  • Representation harms in computer vision
  • Web searches, translation, and language processing online
  • Advertising, recommendation, and misinformation on social media
  • Building, explaining, and auditing fairer algorithms

Learning Outcomes

At the end of this course students will:

  • Critique the strengths and weaknesses of models and inherent values used in machine learning research and technologies
  • Understand how values and human biases are embedded in machine learning algorithms, and critically reflect on power relationships
  • Understand how fairness is theorised in a range disciplines, from social science to computer science, as well as in many practical applications of ML
  • Develop a general understanding of the steps required to audit an algorithmic system and evaluate the impact of human choices made during its development