Covers cutting-edge research related to building moral AI systems. Iowa course listing here.
This is a seminar covering Machine Ethics, a recent and growing subfield of AI/Cognitive Science that is aimed at computationally modeling ethics to 1) better understand ethics and 2) ensure that AI systems are ethical. This is an important task as AI systems are being deployed in risky domains like law, healthcare, and the military. Being a seminar, you will spend time reading and reflecting on cutting-edge technical and philosophical papers.
Our main question(s) will be: How can we formally model ethics (via symbols & rules, vectors, etc.) and should we? Where "formally" here means "described in a way that it can be run as an algorithm and thus used in AI systems." In this context, we will cover (at least) the following topics:
At the end of the semester, students should have the knowledge to explore aspects of machine ethics in their own research and analyze AI techniques in other courses with a critical lens. Specifically, by the end of this course students should:
These are all useful research skills and are sought out by top universities and companies working in AI, law, and more.
There is NO textbook for this course. Each week we will be reading and discussing research papers. A rough calendar for the course is provided below.
| Week | Topic | Overview |
|---|---|---|
| Week 1 | Intro and Course Logistics | Course introduction and motivation |
| Week 2 | Cognitive Modeling / Machine Ethics | A framework for analyzing models / Approaches in machine ethics |
| Week 3 | Moral Philosophy | Prescriptive Ethics: Virtue Ethics, Consequentialism, Deontology |
| Week 4 | Moral Philosophy / Logic | Kantian Moral Theory / Propositional and Predicate logic |
| Week 5 | Deontic Logic | Modal logic / Deontic logic |
| Week 6 | Symbolic Approaches in Machine Ethics | Rule-based AI systems for modeling moral reasoning |
| Week 7 | Symbolic Approaches in Machine Ethics | Rule-based AI systems for modeling moral reasoning |
| Week 8 | Symbolic Approaches in Machine Ethics | Rule-based AI systems for modeling moral reasoning |
| Week 9 | Statistical Approaches in Machine Ethics | Valued Aligned Neural Networks: LLM guard-rails |
| Week 10 | Statistical Approaches in Machine Ethics | Value Aligned Reinforcement Learning: Markov Decision Processes |
| Week 11 | Hybrid Approaches in Machine Ethics | Neuro-Symbolic Models: Bayesian theory / Belief Functions / CP-Nets |
| Week 12 | On the Possibility of Machine Ethics / Final Paper Intro | Analyze the task of machine ethics – is it possible to build a moral machine? |
| Week 13 | On the Possibility of Machine Ethics / Proposal Due | Final paper proposal / Continued analysis of moral machines |
| Week 14 | No Classes | Thanksgiving Break |
| Week 15 | Student Paper Workshop | Discuss paper ideas and writing / Get feedback |
| Week 16 | Student Paper Workshop | Discuss paper ideas and writing / Get feedback |
| Week 17 | Finals Week | No Final Exam / Submit final paper |