- Class Meeting. Mondays and Wednesdays 3:05 – 4:20 pm in LSRC A247.
- Instructor. Brandon Fain. Office hours Tu/Th 1-2 pm on zoom (register for time slot & link here) and Fridays 2-3 pm in-person at LSRC D104.
- Graduate Teaching Assistant. Youngjo Min. Office hours W/F 11 am – 12:30 pm on zoom at duke.zoom.us/j/7813183427
- Course Platforms:
- Canvas learning management system
- Ed Discussion forum for questions and discussion (accessible from Canvas)
- Gradescope submission and grading (accessible from Canvas)
Description
This course examines the impact of moral and ethical thought on research in algorithms, artificial intelligence, and machine learning. Central themes will include bias, fairness, justice, and alignment to human values and norms. Traditional classifiers and risk models will be considered, as well as modern generative AI models for language and images. The class will read extensively in the interdisciplinary fairness and ethics for algorithms and AI research communities with a particular focus on the ACM FAccT and ACM/AAAI AIES conferences. Students will conduct original research projects designed to have the potential for publication in these same venues.
Topics
Appropriate topics for student research are those that align with the aforementioned conferences. For example, AIES lists the following illustrative topics of interest.
- Trustworthy AI systems
- Governance, regulation, control, safety, and security of AI
- Value alignment and moral decision making
- Interpretability, explainability, and transparency of AI
- Fairness, equity, and equality in AI
- Human-centered AI, human-AI interaction, and human-AI teaming
- Ethical models/frameworks around AI and data
- AI, lawmaking and the judiciary
- AI in public administration, social service provision, and social good
- AI, surveillance, and privacy
- AI, markets and competition
- AI, health, and wellbeing
- AI and creativity, literature and the arts
- AI, democracy and social movements
- Cultural, geopolitical, economic, employment, and other societal impacts of AI
- Environmental costs and climate impacts of AI
Background and Prerequisites
Students should have some prior experience with machine learning. Additional experience with independent research; deep learning; algorithms; and moral, ethical, and human factors will all be helpful to get the most out of the course. There are no enforced prerequisites; speak with the instructor if you have questions.
References
The schedule presents all required readings, mostly research and perspective articles. The following optional texts are recommended for those new to the field who would like more comprehensive resources for foundational learning and reference.
