SPI 352/COS 352: Artificial Intelligence, Law, & Public Policy
PRINCETON UNIVERSITY, FALL 2024
Location: Robertson Hall 016
Time: Thursdays 1:30pm-3:20pm
Instructor
Prof. Peter Henderson, J.D., Ph.D.
Assistant Professor in CS/SPIA
Preceptor
PhD Candidate SPIA
Course Description
This course examines the implications of artificial intelligence (AI), particularly foundation models, for law and public policy. We will cover how AI affects and reshapes legal doctrine and policy, including: intellectual property law, administrative law, anti-discrimination law, and more. Also covered will be emerging regulatory policies and legislative efforts around AI, as well as the limits of proposed approaches. Emphasis will be placed juxtaposing the legal and policy considerations with technical design decisions, in an interdisciplinary and accessible way. This course is suitable for students of all backgrounds; no technical knowledge is assumed.
Course Expectations & Grading
Components
- đź“ť Reading Responses (25%): Due 5pm ET the day before class on Canvas
- đź‘Ą Participation (20%): Attending lectures and precepts
- đź“„ Final Paper/Project (55%):
- Literature Review & Project Proposal - October 18th (10%)
- First Draft - November 25th (15%)
- Final Draft - December 13th (30%)
Course Schedule
| DATE | TOPIC | LECTURE AGENDA |
|---|---|---|
| 9/5 | Preliminaries: What is AI? |
What kinds of AI do you immediately think of when you think "AI regulation?" Do a search on the web, what is your preferred definition of AI, and why? Do you think we even need to define AI in regulatory efforts at all, why or why not? Required Readings:No required reading for first day of class. Optional Readings:
|
| 9/12 | Intellectual Property I: Copyright |
Should model training be a fair use? What if the model outputs exact pieces of the training data? How much verbatim regurgitation by models should be acceptable? Do you think the courts should resolve this or should Congress step in, why or why not? Should humans have authorship rights in AI-generated content purely from prompting? Required Readings:
|
| 9/19 | Intellectual Property II: Right of Publicity + Privacy I |
How do you think we should regulate use of likeness? What does it mean for a voice clone or character to be too close in likeness to a real person? What if multiple people have similar voices? Should restrictions on use of likeness expire ever? What about elected officials, should we be more or less restrictive about use of their likeness? Required Readings:
|
| 9/26 | Tort Liability & Section 230 |
Should large language models be immune from liability under Section 230? What about recommendation systems? Where should we draw the line? What is considered reasonable care under a negligence standard in tort law? Required Readings:
|
| 10/3 | Free Speech and First Amendment |
As the conflicting readings suggest, there is significant grey area around the applicability of the First Amendment, what do you think the right position is? What makes you think this is the right position and how do you assess the "correctness" of your position? What are the consequences of taking one position of the other? Required Readings:
|
| 10/10 | National Security Threats & Uses, Export Controls, & The Executive's Power |
What are the pros and cons of a distributed state-level approach to AI regulation versus an approach relying on executive powers? What about comparing a pre-clearance regime to a post-deployment monitoring regime? How should we balance expanding national security powers around AI against containing the risks of AI? How does the first amendment interact with executive powers? Required Readings:
|
| 10/17 | Fall Break | |
| 10/24 | The Administrative State I: Antidiscrimination Law |
What is your assessment of the complications of determining whether Facebook / Meta violated the Fair Housing Act? Do you agree with the resolution of the case? Do you think WorkDay should be liable for discrimination or should it be a problem only for direct employers? What if we go further upstream to OpenAI and other foundation model providers if they power WorkDay's service? Required Readings:
|
| 10/31 | The Administrative State II: Antitrust and Labor |
What tools can we leverage to constrain concentration of power under increased automation? What role should different areas of law play in this? Should we battle concentration of power at all? Do you think labor organizing will be effective for this? What about antitrust law? Required Readings: |
| 11/7 | The Administrative State III: Challenges for Regulators in the Administrative State |
Do you think the administrative state remains a viable option after recent supreme court jurisprudence? Is regulation better off in the hands of the states, or the executive? Required Readings:
|
| 11/14 | The Administrative State IV: Uses in the Administrative States and Constraints on Itself |
Does it matter if a human is in the loop? If so, when and how can we make sure that humans are actually doing their job when they're in the loop? If not, why not? Are there government uses of AI that you're more or less comfortable with? Required Readings:
|
| 11/21 | Project Discussions | |
| 11/28 | Thanksgiving Recess | 12/5 | Global Approaches and Comparing Regulatory Methods: EU AI Act, Regulatory Efforts in China, International Agreements |
How do the different regulatory approaches compare? Which one do you think is better and why? Required Readings:
|