Human vs. Algorithmic-Decision Making:

Bias, Fairness & Transparency

Seminar (7 CP)

Winter Semester 2015-16

Instructor: Krishna Gummadi

Teaching Assistants: Oana Goga, Juhi Kulshrestha, Muhammad Bilal Zafar

News

• Class time changed to Friday 14:00 to 16:00.
• Seminar kickoff meeting takes place on October 23, 2015 (Friday) at 10:00 in room 005, E 1 5.

Topic

The concepts of bias, fairness and transparency have been studied previously in philosophical, social, cultural, economic and legal frameworks. However, as automated data analysis replaces human supervision and intuition in decision making, and the scale of the data analyzed becomes “big”, there is a growing need for incorporating these concepts into algorithmic decision-making frameworks [1, 2, 3]. While these concepts are somewhat intuitive, quantifying them in algorithmic frameworks is a non-trivial task. For example, what does it mean for an algorithm to be fair or unbiased? Or, how do we quantify the extent to which an algorithm is discriminating against subjects of a protected class, and so on.

In this seminar, we will investigate some of these questions. We will study how these concepts are defined and understood in humanities and social sciences and try to map them to algorithmic (data mining and machine learning) frameworks.

Prerequisites

Prerequisites for the course include undergraduate courses on algorithms, programming languages, and information systems.Students are also expected to be familiar with material covered in advanced undergraduate / master-level courses in data mining or machine learning.

The course projects would require significant amount of programming effort towards analyzing data or towards implementing and deploying new system designs. Experience in one or more of the programming language such as Java, C/C++, Python is required.

Signup

Students interested in participating in the seminar should register by October 30, 2015 using the signup form.

Organization

When: regular meetings every Friday from 14:00 to 16:00.

The first meeting is on October 23.

Where: room 005, E1 5 (MPI-SWS building, UdS Campus).

Office hours: Tuesdays 14:00 to 16:00 in room 302, E 1 5.

Format

Each week we will read one paper from the selected topics. Students do not have to present papers, instead one of us (TAs/instructors) will present the paper to initiate the discussion. The course will consist of three components:

• Paper discussion: The audience for each presentation should actively participate in the paper discussion. A seminar course thrives on lively discussion.

• Submitting paper reviews: Before the class each week, all students have to submit a review for the paper to an online portal. Students can view each other's reviews and rate them. For more details on how to write a good review, see here.

• Research project: While the papers that we read will provide you with sufficient background knowledge to understand the various research problems in this space, we will also have a course project to provide a more hands-on learning experience. Details of the project will be posted in mid-November. Instructors will assist the students in selecting and defining the scope of the project.

Paper Reading List

List of covered papers and the lecture slides are posted here.

Course Project

Details of the course project can be found here.

Grading