Summary This set of assignments provides students with the opportunity to read about and reflect on the many different ways in which bias can occur in AI and ML problems. Students will learn about the technical definition of bias, how this relates to the more colloquial version, why it is so difficult to define what it means to be fair, and the importance of understanding what biases exist in real-world data sets. Also included are strategies and resources for class discussion, connecting to technical topics, and a discussion of specifications grading as an approach for assigning written work.

Topics Ethics and broader impact, machine learning

Audience College students of any major

Difficulty Medium - 2-3 hours needed outside of class.

Strengths Encourages students to make connections between technical issues and real-world impact; helps them understand the importance of careful problem design, explains why systems can have unintended effects.

Weaknesses No coding involved. Future work could include a stronger connection to theoretical issues in statistics and learning

Dependencies None.

Variants I do this with other topics as well, including open source/walled garden, autonomous weapons, COVID, contact tracing and privacy, and computational art. There are no shortage of interesting issues for students to reflect on.

README.pdf

Bias on the Web.pdf
COMPAS.pdf
Measuring Diversity.pdf
ML Bias.pdf

Summary	This set of assignments provides students with the opportunity to read about and reflect on the many different ways in which bias can occur in AI and ML problems. Students will learn about the technical definition of bias, how this relates to the more colloquial version, why it is so difficult to define what it means to be fair, and the importance of understanding what biases exist in real-world data sets. Also included are strategies and resources for class discussion, connecting to technical topics, and a discussion of specifications grading as an approach for assigning written work.
Topics	Ethics and broader impact, machine learning
Audience	College students of any major
Difficulty	Medium - 2-3 hours needed outside of class.
Strengths	Encourages students to make connections between technical issues and real-world impact; helps them understand the importance of careful problem design, explains why systems can have unintended effects.
Weaknesses	No coding involved. Future work could include a stronger connection to theoretical issues in statistics and learning
Dependencies	None.
Variants	I do this with other topics as well, including open source/walled garden, autonomous weapons, COVID, contact tracing and privacy, and computational art. There are no shortage of interesting issues for students to reflect on.