CS 5/692, Winter 2020

Ethics in AI and Machine Learning Research

Overview

Computer science research has changed dramatically over the last ten years, both in terms of the ways in which it is conducted as well as the ends to which it is applied. Research in our field is powered by large quantities of data- Tweets, clicks, records, posts, etc.- which, as a rule, were all created by somebody, somewhere. Furthermore, the algorithms and computational techniques that we develop are finding their way into every corner of our lives. They determine what news we see, affect our financial and professional choices, and are widely used as part of our criminal justice system. As they do so, they interact with all aspects of our society, flattening some forms of inequality while amplifying others, often in subtle and surprising ways. Seemingly-minor methodological choices by system designers can have profound consequences. This course will explore these and other issues, with the goal of preparing researchers-in-training with the knowledge they need to responsibly conduct research in this area, and also to prepare them for their professional careers.

Topics will include traditional RCR subjects such as informed consent, ethical reasoning, conflicts of interest, etc., but will address these issues through the lens of modern machine learning and “big data” problems. Additional topics will include issues of algorithmic bias, anonymity, data privacy, value-sensitive design, fairness/accountability/transparency, etc.

Textbooks & Resources

• Course Page on OHSU Sakai
• Cathy O’Neil, Weapons of Math Destruction
• Sara Baase & Timothy M. Henry, Gift of Fire, 5ed
• Nicholas Steneck, ORI Introduction to the Responsible Conduct of Research (available online)

Course goals

By the end of the course, students will be:

• Conversant with standard frameworks of ethical reasoning (virtue, deontology, etc.)
• Familiar with the history of the use of humans as subjects in scientific research, and with modern norms thereof (Belmont Report, NIH Common Rule, etc.)
• Able to translate and apply concepts such as informed consent, etc. to machine learning research contexts
• Familiar with current thinking in the Fairness, Accountability, and Transparency in Machine Learning movement
• Prepared to identify and evaluate potential areas of ethical concern in their work, both as researchers in and practitioners of machine learning.
• Aware of modern intersectional models of gender, racial, socioeconomic, etc. diversity and their interaction with machine learning and natural language processing

Basic course structure

Week 1–5: Principles of ethical reasoning; policy; human subjects & privacy; design principles (value-sensitive design); case study methodology;

Week 6–8: Algorithms, applications, and bias; fairness, accountability, transparency

Week 9-11: dataset documentation? handling sensitive data? professional ethics in CS/AI/ML; ethical codes;

Logistics

When & Where?

Mondays, 14:00 in CDRC 3200

Starts: January 6

Ends: March 13

Who?

Instructor: Steven Bedrick

Office Location: Gaines Hall, 21

Office Hours: Wednesday, 9:00-10:30 AM