The goal of this practicum is to integrate methods from natural language processing with social psychological perspectives on race to build practical systems that address significant societal issues. Readings will be drawn broadly from across the social sciences and computer science. We will discuss how to work with large, complex datasets and highlight the benefits of working directly with practitioners to tackle some of society’s most vexing social problems. Prerequisite: CS224N, PSYCH290, or equivalent background in natural language processing. Students interested in participating should complete the online application for permission below. Limited enrollment.
Tuesdays 1:30 - 4:00 pm, McMurtry Art & Art History Building, Room 350
Jennifer L. Eberhardt
Email: jleberhardt@stanford.edu
Office Hours: Mondays 12 – 1 pm, or by appointment
Dan Jurafsky
Email: jurafsky@stanford.edu
Class Office Hours: Tuesdays 4:00 – 5:00 pm, or by appointment
Hodan Farah
Email: hfarah@stanford.edu
Office Hours: Thursdays 1- 2 pm, @ Coupa Cafe at Y2E2
Maggie Harrington
Email: mperry3@stanford.edu
Office Hours: Mondays 2:30 - 3:30 pm, or by appointment @ Building 420, Room 382
Course Description: This course is an interdisciplinary practicum designed to get students with computational backgrounds and students with social science backgrounds to collaborate on building practical systems that address societal issues related to race and inequity. Readings will be drawn broadly from across the social sciences and computer science. We will discuss how to work with large, complex datasets and highlight the benefits of working directly with practitioners to tackle some of society’s most vexing social problems. Prerequisite: CS224N, PSYCH290, or equivalent background in natural language processing. Students interested in participating should complete the online application for permission at https://web.stanford.edu/class/cs329r/. Limited enrollment.
| Reaction Papers: | 20% of final grade |
| Class Participation: | 10% “ “ |
| Discussion Leadership: | 10% “ “ |
| Final Project: | 60% “ “ |
| Discussion Questions: | Mondays before 5:00 pm (each week) |
| Reaction Papers: | Mondays before 5:00 pm (two times during quarter) |
| Project Proposal: | Oct 10 midnight |
| Rough Draft of Project: | Nov 14 midnight |
| Project Presentation: | Dec 5 in class |
| Final Project Writeup | Dec 12 midnight |
| Sep 26 | Introduction (including topics, requirements, and the value of community partnerships) |
| Oct 3 | The Transmission of Bias and the Mechanics of Inequality |
| Oct 10 | How We Police |
| Oct 17 | How We Teach (and conclusion of How We Police) |
| Oct 24 | How We Talk: A Focus on Social Media |
| Oct 31 | How We Talk: A Focus on Politics |
| Nov 7 | NO CLASS – Democracy Day |
| Nov 14 | How We Work |
| Nov 21 | NO CLASS – Thanksgiving Recess |
| Nov 28 | How We Rebuild |
| Dec 5 | Presentations of Final Projects |
It is important that you attend each session and complete the readings prior to class. The discussion and interaction during class time will be an integral part of the course.
Throughout, we will engage challenging questions about human behavior and society that intersect with our own lives. It is especially important that we all remain open to the expression of views that differ from our own and that we are willing to reconsider our own views. Maintaining humility about the rightness of one’s own views and appreciation for the insights of others are both essential to our being able to learn together. Of course, even as we challenge each other’s ideas and arguments, we want to maintain an atmosphere of respect and collegiality.
You will be required to write two short reaction papers during the quarter. These papers should be approximately 2 pages (double spaced). The papers can be written for any week of your choosing, as long as they are posted to Canvas the day before the class meets to discuss the topic/readings about which you have written. These papers should not be descriptive summaries of the readings. Instead, they should offer a critical analysis. For example, you may choose to discuss a problem or limitation with one of the readings (or with several of the readings) and offer a better approach or method. You may propose a specific study to conduct. You may discuss a theme that seems to cut across all of the readings. You may propose a new theoretical framework for understanding a phenomenon discussed in the readings. There are many options. Regardless of the option you choose, you should strive to lay out a coherent, well-defended argument.
Each week you will be asked to generate at least two discussion questions based on the readings. These questions will be factored into your participation grade. The questions will help us to understand common points of interest. And we will use them to help guide the discussion in class. Again, these questions should not be descriptive. Instead, they should be probing, analytical, and thought-provoking. Indeed, the quality of our discussions will critically depend on your contributions in this regard. Your questions should be posted on Canvas no later than 5 PM each Monday.
Once during the quarter, you will be asked to help lead the discussion for the week. You are expected to meet with your group partner(s) beforehand to agree on the questions and issues you will use to frame and guide the discussion. You should be prepared to point out big themes. We will rely on you to place the research in context. What is the value of the work? Why does this work matter to the fields of linguistics, computer science, psychology, or to the public at large? Does the research have policy implications worth exploring? Bring in your own expertise on the issues to help inform the class and to push us beyond the readings.
The final project is a chance to apply NLP to a societal issue related to race. During the course of the quarter, we will discuss projects that developed from partnerships with leaders in industry, schools, and police departments (with whom we have built ties over the years) as well as repos of social media data. The projects you propose could involve using text data to conduct analytic studies of race and inequality in partner domains, testing NLP-powered interventions like content moderation or training, or building other practical tools. For areas where data requires IRBs or long-term contracting, the project can be a proof-of-concept proposal, with an implementation applied to pilot or sample data. The projects can be done individually or with a partner; if with a partner, we recommend a cross-disciplinary mix of computer scientists and social scientists. Project milestones include:
