CS 5382 - Practical Principles for Designing Fair Algorithms

Spring. 3 credits. Student option grading.

Students should have experience coding in Python and have taken at least one introductory course in machine learning or data science. Offered in New York City at Cornell Tech.

E. Pierson.

Algorithms increasingly guide high-stakes decision-making across many domains. This has potential upsides, since algorithms can improve decision-making, but also serious risks, since recent years have showcased the many ways that algorithms can be biased. This course will teach you principles for designing fair algorithms, emphasizing accessibility to a broad audience via practical takeaways which are directly relevant to the real world through case studies and guest speakers. Case studies will be drawn from diverse settings where algorithms are applied, such as large language models, speech recognition systems, healthcare, criminal justice, sustainability, and education. Students will come away with a strong understanding of how algorithm-related choices can have widespread societal impact.

Outcome 1: Write code in Python to computationally demonstrate biases in end-to-end algorithmic systems based on choices of data, variables, modeling, and outcomes.

Outcome 2: Apply mathematical definitions of fairness to real-world case studies to explain decisions made by both humans and algorithms.

Outcome 3: Enumerate challenges to practitioners in algorithmic-guided decision-making (including feedback loops, interpretability, and strategic behavior) and explain how these challenges can lead to broader societal impacts.