Computational Ethics for NLP

CMU CS 11830, Spring 2020

T/Th 10:30-11:50am, SH 214

Yulia Tsvetkov (office hours by appointment),
Alan W Black (office hours: Wednesdays 12-1pm, Zoom link on Piazza),
TA: Anjalie Field (office hours by appointment),
TA: Michael Miller Yoder (office hours by appointment),

HW 4: Design a Computational Ethics Homework Assignment

Due 11:59pm, Thursday 4/16

Submission: Submit a zip/tar file to Canvas. Include your write-up (titled Lastname_Firstname.pdf) and a zip/tar folder containing your code. Code will not be graded.


This class is designed to get you thinking about the ethical implications of NLP and machine learning systems you build. Most NLP and CS students don’t take ethics classes; we want you to think about how you might share concepts you learned in this course even after it finishes. In this assignment, you will demonstrate this critical thinking by designing a homework assignment for the course. You do not have to do the project you describe, but just design it. The goal is to go beyond just modeling approaches and develop a framework for addressing/identifying who benefits or could be harmed from an NLP system, unintended consequences and biases of NLP systems, or using NLP to analyze injustices in society or for social good.


We’re turning the tables for this assignment: you are an instructor for this class and are designing a homework assignment for students. Again, you do not have to do the project you describe, but just design it.


First, identify which unit in the course your assignment will be designed around. The units are:

Your assignment should include learning goals, an overview, specification of a dataset (that currently exists), illustrative examples from that dataset, basic requirements, and advanced analysis. For the goals and overview, think about what you want students to get out of your assignment (this is not trivial). The basic requirements should indicate a dataset and a baseline machine learning model or quantitative analysis. In advanced analysis, list one or more options for novel modeling or analysis, ideally that hasn’t been published before.

Your proposed assignment should be sufficiently different from existing assignments and projects in the course. It should also be simple enough to be completed by students in about the same amount of time that you spend on the homework assignment.

For inspiration, look to past lectures, the reading list, project ideas list, project ideas you considered but dropped for your class project, current events, and questions/tasks from your own field. Below are a few specific areas and datasets you can consider (remember to include examples from these datasets):

Write-up: The assignment description should be 1-2 pages in ACL format, FirstName_LastName_hw4.pdf. Please do not write more than 3 pages. Your proposed assignment write-up should include:

See the structure of any of the other homework assignments for guidance. You do not have to specify what you are looking for in a write-up or grading.

Grading (100 points)