Machine Learning, Ethics, and Society

Course Description

Overview. The practice of Machine Learning (ML) increasingly involves making choices that impact real people and society at large. This course covers an array of ethical, societal, and policy considerations in applying ML tools to high-stakes domains, such as employment, education, lending, criminal justice, medicine, and beyond. We will discuss:

1. the pathways through which ML can lead to or amplify problematic decision-making practices (e.g., those exhibiting discrimination, inscrutability, invasion of privacy, and beyond);
2. recent technological methods and remedies to capture and alleviate these concerns; and
3. the scope of applicability and limitations of technological remedies in the context of several contemporary application domains.

The course's primary goals are: (a) to raise awareness about the social, ethical, and policy implications of ML, and (b) to prepare students to critically analyze these issues as they emerge in the ever-expanding use of ML in socially consequential domains. Topics that will be covered include:

• Unfairness, bias, and discrimination through ML
• Transparency and explainability for ML
• The societal aspects of ML safety, robustness, and causality
• Privacy, profiling, and surveillance via ML
• Accountability and governance of ML-based systems
• Long-term social and economic implications of AI and ML (long-term population effects; labor market effects; and trust)
• Human factors and participatory designs
• Applications domains and policy considerations (e.g., misinformation and polarization in online platforms; diversity in tech)
• Misc. topics in AI ethics (e.g., ethics of robots and self-driving cars)

Prerequisite knowledge. Background in Machine Learning is a prerequisite. In particular, students are expected to have adequate knowledge of probability, statistics, linear algebra, and optimization as it pertains to standard ML topics. Background in algorithm design, mechanism design and game theory, and coding skills will be helpful but not necessary.

Performance Assessment. The final grade will be computed based on:

• written mid-term and final exam
• homework assignments
• class project (written report + class presentation)
• participation in class discussions

For a more detailed syllabus, please view this document. (Note: you need to be signed in via your CMU credentials to be able to access the document).

---