CSE 543 is the new permanent course number for Ethics for AI and Robotics, cross-listed as ROB 543. This is also the first course with a "CSE" course number, rather than "EECS". It shows up in the Registrar's list at the beginning of the EECS courses, but some people may overlook it. Please help people find it if they might be interested.
NOTICE: If you need permission of the instructor (and all undergraduates do), send me email (kuipers@umich.edu) explaining why you should take the class.
Artificial Intelligence, Machine Learning, and Robotics have seen dramatic progress in the last several decades. There is increasing excitement and apprehension about the impact of these technologies, deployed in our world and our human society.
Ethics is the discipline within philosophy that considers which actions we humans see as right or wrong, or as good or bad. As we design intelligent artifacts that make their own decisions about how to act, and as they act within the human world, we ask how we can ensure that they will act ethically.
Two important questions arise.
First, like any other powerful technology (e.g. nuclear power, genetic engineering), there are important ethical questions about how AI and robotics technology can and should be deployed, and what its impact will be on society. This topic includes regulations, and the processes by which regulations are proposed, adopted, and enforced.
Second, unlike other technologies, AI (and thus intelligent robotics) involves creating agents that make their own decisions about how to act in the world. Ethics is a kind of foundational knowledge that humans use to decide how to act. We need to understand the structure of that knowledge, so the AIs we create will have the knowledge they need to act appropriately.
Do we mean that humans must be ethical as we design and deploy intelligent systems? Do we mean that the systems we design and deploy must be capable of deciding what is ethical for them to do? Most likely, the answers to both questions will turn out to be “Yes!” The follow-on question is “How do we do that?”
The semester will be organized around several major topic areas:
In the course of discussing research on these problem areas, we will draw on concepts from philosophical ethics, and from engineering design, law, economics, evolution, history, human development, etc.
An important question for researchers in artificial intelligence and robotics is how the knowledge relevant to making ethical decisions can be represented computationally in a knowledge base, and how it can be acquired.
Lectures: two each week, 80 minutes each (MW 4:30 - 5:50 pm). Discussion: one each week, 50 minutes (F 3:30 - 4:20 pm). (Attendance will be taken, and counted toward the Participation grade.)
We plan to have a number of guest lecturers.
The lectures will include opportunities for class discussion. The Discussion meeting will go into further depth on issues that come up in lectures, but may also explore important issues in current news. Some discussions will involve the entire class, but we may also break into small groups.
What if you must miss attending class?
Attendance does count toward the Participation grade (and so does interaction on Piazza). A modest number of absences are inevitable and will not be penalized. Large numbers of absences are a problem, though.
Covid
Current University of Michigan policy requires all classes to be held in person, and all students, faculty, and staff to be fully vaccinated. Furthermore, the ventilation of all classrooms has been upgraded.
Each student will attend the classes, do the readings, participate in the discussions, and write two papers. Attendance and participation will have significant weight (20%) in the course grade.
There will be extensive reading assignments, typically two required papers for each class. (Additional recommended readings will often be provided.) For each required paper, before the corresponding class, enter a thoughtful comment on the Piazza thread provided for that paper. These comments will be counted toward the participation grade.
Pick a topic for your two term papers by the end of January (2-5-2024). Don't restrict yourself to what we've already covered in the course. Look ahead in the schedule and the reading assignments, or come up with something else. Submit for comments and advice, but not counted toward the course grade.
In the first paper, due at mid-term (3-4-2024), you will review the available literature related to the topic you have selected. The goal of your paper is to identify, clarify, and summarize the major positions on your topic. Counts for 40% of the course grade.
In the second paper, due at the end of the term (4-22-2024), you will pick a question, take a position on how it should be answered, and justify your position, responding clearly to anticipated arguments from critics of that position. Counts for 40% of the course grade. Focusing either or both papers on the topic submitted in January is not required, but it will obviously make both papers stronger.
Participation counts for 20% of the grade. This includes attendance in lecture and discussion sections (6%); participation in class (6%); posting on Piazza your comments on the readings (7%); and submitting the final Course Evaluation (1%).
Course grades will be curved if needed, though in recent years no curve has been necessary.
You will need to find and read additional papers and possibly books as part of your literature review and final term paper. Be sure that you know how to use Google Scholar and the UM Library's Online Journal collection for tracking down references.
The following paper is a good example of a literature review, and it is also highly relevant to the content of the course. Read it before the start of the course.
When you write your own literature review, it need not match this in length or style, but this paper is an aspirational target. For your paper, imagine that you are providing detailed help on a particular focused topic, to a friend who wants to get started doing research on that specific topic.
This course now has a permanent course number: CSE 543, cross-listed as ROB 543.
These are graduate course numbers, but undergraduates may request instructor's permission to enroll. Email me at kuipers@umich.edu describing your relevant background and your motivation for taking the course.
This course describes and discusses the ethical issues raised by AI and Robotics, reading and analyzing arguments from a number of disciplines, identifying and posing ethical questions, evaluating potential solutions, and formulating future research questions. The course will include guest lectures from experts in computer science, philosophy, cognitive science, psychology, public policy, law, etc.
The two papers should demonstrate that you can search, find, and review good quality references beyond those handed out in class, and that you can put your own creative and critical insights into formulating a good problem and deeply and thoughtfully exploring solutions to it, anticipating and responding in advance to criticism.
The graduate course has been approved to satisfy the following CSE Graduate Program requirements: depth (not cognate) requirement for the CSE PhD, and the 500-level and technical elective requirements for the CSE MS (confirmed: 10/30/2023).
For undergraduates, CSE 543 can substitute for the EECS 496 requirement (confirmed 11/16/2023). It no longer satisfies the CoE Intellectual Breadth requirement.
We use Piazza as our platform for online discussion outside of class. The system is designed to get you help quickly and efficiently from classmates, the TA, and myself. Rather than emailing questions to the teaching staff, I encourage you to post your questions on Piazza. If you have any problems or feedback for the developers, email team@piazza.com.
Find our class page at: https://piazza.com/umich/winter2024/ethicsforaiandroboticsw24/info
Each lecture has two (very occasionally more) assigned readings, and several suggested readings. Each student should write a short comment on each assigned reading, tagging it with the Piazza folder "reading_comments", and the subfolder corresponding to the lecture number (L1-L27). (The lecture numbers are on the schedule of lectures.) Feel free to add messages discussing (respectfully) other people's posts. Your posts count toward your Participation grade.
Students learn more when they collaborate actively with each other. But you also learn more when you take responsibility for doing your own work. Be prepared to teach each other, and to learn from each other, but not to do each others' work.
Discuss all the topics in the course, especially the topics of your papers, with each other. Recommend papers to each other. Read and comment on each others' drafts as alpha and beta readers. But formulate and defend your own opinions, and write your own papers.
It is not a course requirement that you publish your paper in a conference or journal, but that's certainly a great outcome! After the course is over, it is entirely reasonable to combine and revise closely related papers into a joint submission for publication.
Anyone working in this area should understand the lesson of this children's poem.
Nonzero describes human biological and cultural evolution in a framework provided by game theory.
My chapter in the Oxford Handbook of Ethics of AI (2020) and my articles in Frontiers in Robotics and AI (2022) and CACM (2023) provide overviews of some of the topics we will cover in the course.
Then read about the problem of lethal autonomous weapons.
Links to other resources and recommended books.