Teaching

DS-GA 3001: (Gen)AI Accountability on the Ground: Harms, Law, and Practice

Description

This is a multidisciplinary course that explores the technical and legal dimensions of harms caused by both traditional and generative AI systems. We draw both from academic research and real-world case studies in domains like finance, housing, healthcare, and employment to understand how (Gen)AI harms happen, and can be mitigated, on the ground–i.e. in a way that is practically applicable.

Students will learn to identify unfair, opaque, or otherwise problematic behaviors in AI and GenAI systems, apply practical techniques to mitigate harms such as discrimination, privacy violations, and arbitrariness, and learn when and which discrimination, privacy, and other laws apply to algorithmic systems and how to be compliant with these laws.

Prerequisites

There are no formal prerequisite requirements, but I assume knowledge of AI/ML concepts from an undergraduate or graduate class (i.e., words like loss function, optimization, input features, classification, correlation, gradient, regression, random forest, and hyperparameter should be familiar to you) and optional homework assignments assume knowledge of python and scikit-learn, numpy etc. If you don’t have this background, I believe there’s a way to still take the class, but you should email me first.

Grading

• 15% Participation
  • Please come to class ready to discuss the readings assigned for that day (the readings listed on that day). Part of this class will be discussion-based, so we’ll be talking about the readings together.
• 15% Quizzes
  • Many classes will have an in-person paper quiz given about the (required) reading. It will be very easy, this is just to make sure you have done the reading for the class.
• 35% Midterm
  • In-class, closed book but with a cheat sheet where you can write anything you want (double sided). It will cover the materials learned in approximately the first 2/3 of class.
• 35% Project (15% Presentation, 20% Final Report)
  • In pairs, you will come up with a unique research idea to study inspired by topics we speak about in the class. Students get to decide if they want to try to create a viable research paper based on the project, or simply create a class project to fulfill learning objectives. This doesn’t impact grading, but students in research groups will have more office hours available to develop their ideas.