263-5215-00L  Fairness, Explainability, and Accountability for Machine Learning

SemesterSpring Semester 2019
LecturersH. Heidari
Periodicitynon-recurring course
Language of instructionEnglish
CommentNumber of participants limited to 40.

The deadline for deregistering expires at the end of the second week of the semester. Students who are still registered after that date, but do not attend the course, will officially fail the course.


Abstract
Objective- Familiarize students with the ethical implications of applying Big Data and ML tools to socially-sensitive domains; teach them to think critically about these issues.
- Overview the long-established philosophical, sociological, and economic literature on these subjects.
- Provide students with a tool-box of technical solutions for addressing - at least partially - the ethical and societal issues of ML and Big data.
ContentAs ML continues to advance and make its way into different aspects of modern life, both the designers and users of the technology need to think seriously about its impact on individuals and society. We will study some of the ethical implications of applying ML tools to socially sensitive domains, such as employment, education, credit ledning, and criminal justice. We will discuss at length what it means for an algorithm to be fair; who should be held responsible when algorithmic decisions negatively impacts certain demographic groups or individuals; and last but not least, how algorithmic decisions can be explained to a non-technical audience. Throughout the course, we will focus on technical solutions that have been recently proposed by the ML community to tackle the above issues. We will critically discuss the advantages and shortcomings of these proposals in comparison with non-technical alternatives.
Prerequisites / NoticeStudents are expected to sufficient knowledge of ML (i.e. they must have taken the "Introduction to Machine Learning" or an equivalent course).