Note: For an updated resource, please see fairmlbook.org.

CS 294: Fairness in Machine Learning

UC Berkeley, Fall 2017
Time: Monday and Friday 2:30PM - 3:59PM
Location: Soda 405
Instructor: Moritz Hardt

Grading policy: 50% in class participation, 50% project

Undergraduate enrollment policy: With permission from the instructor only. Email transcript, and description of any relevant prior projects or research experience.

Students will receive access to a slack channel for class related discussions.

Updates

See these NIPS 2017 tutorial slides and video for a two-hour introduction to the topic.
Important: Projects will be presented on Tuesday, Dec 12 from 11:30–2:30.
8/15/2017 — The class has reached its enrollment cap. Additional participants are waitlisted for now. If you would like to participate and couldn’t enroll, please come to the first day of class and talk to me.

Summary

This is an intensive graduate seminar on fairness in machine learning. The focus is on understanding and mitigating discrimination based on sensitive characteristics, such as, gender, race, religion, physical ability, and sexual orientation. Recent years have shown that unintended discrimination arises naturally and frequently in the use of machine learning and algorithmic decision making.

We will work systematically towards a technical understanding of this problem mindful of its social and legal context.

Lectures

Day 1 (8/25) Overview talk
Day 2 (8/28) Barocas and Selbst
Day 3 (9/1) Blasts from the past
Day 4 (9/8) Intro to oboservational measures
Day 5 (9/11) Discussing some project ideas
Day 6 (9/15) Guest lecture by Kristian Lum on recidivism prediction
Day 7 (9/18) Fairness through Awareness
Day 8 (9/22) Simpson’s paradox
Day 9 (9/25) First few chapters of Pearl textbook
Day 10 (9/29) Chapter 4, 5 in Peters textbook
Day 11 (10/2) Chapter 6 in Peters textbook
Day 12 (10/6) Chapter 7, 8 in Peters textbook
Day 13 (10/9) Causal fairness papers
Day 14 (10/13) More causal fairness papers
Day 15 (10/16) Causality recap and outlook
Day 16 (10/20) Measurement
Day 17 (10/24) Hand Chapter 2
Day 18 (10/27) Hand Chapter 3
Day 19 (10/31) Hand Chapter 4, failure points of measurement
Day 20 (11/3) Raw Data is an Oxymoron
Day 21 (11/6) Discussing project ideas
No class on 11/10 due to Veteran’s Day.
Day 22 (11/13) Introduction to sampling
Day 23 (11/17) Closer look at crime surveys and measurement
Day 24 (11/20) Critiques of algorithmic decision making
No class on 11/24 due to Thanksgiving.
Day 25 (11/27)
Day 26 (11/30)
Day 27 (12/4)
Day 28 (12/8)
Day 29 (12/10) Project presentations, special time 11:30–2:30

Outline (preliminary)

Last updated: August 17, 2017

Sources of unfairness
Statistical measures of discrimination
Trade-offs and impossibility results
Beyond observational measures
Measurement, sampling
Legal and policy perspectives

Sources of unfairness

Big Data: A Report on Algorithmic Systems, Opportunity, and Civil Rights
The White House. 2016.

Bias in computer systems
Batya Friedman, Helen Nissenbaum. 1996

The Hidden Biases in Big Data
Kate Crawford. 2013.

Big Data’s Disparate Impact
Solon Barocas, Andrew Selbst. 2014.

Blog post: How big data is unfair
Moritz Hardt. 2014

Semantics derived automatically from language corpora contain human-like biases
Aylin Caliskan, Joanna J. Bryson, Arvind Narayanan

Statistical measures of discrimination

Sex Bias in Graduate Admissions: Data from Berkeley
P. J. Bickel, E. A. Hammel, J. W. O’Connell. 1975.

Simpson’s paradox
Pearl (Chapter 6)
Tech report

Certifying and removing disparate impact
Michael Feldman, Sorelle Friedler, John Moeller, Carlos Scheidegger, Suresh Venkatasubramanian

Equality of Opportunity in Supervised Learning
Moritz Hardt, Eric Price, Nathan Srebro. 2016.

Blog post: Approaching fairness in machine learning
Moritz Hardt. 2016.

COMPAS and criminal justice

Machine Bias
Julia Angwin, Jeff Larson, Surya Mattu and Lauren Kirchner, ProPublica
Code review:
github.com/probublica/compas-analysis
github.com/adebayoj/fairml

COMPAS Risk Scales: Demonstrating Accuracy Equity and Predictive Parity
Northpointe Inc.

Fairness in Criminal Justice Risk Assessments: The State of the Art
Richard Berk, Hoda Heidari, Shahin Jabbari, Michael Kearns, Aaron Roth. 2017.

Courts and Predictive Algorithms
Angèle Christin, Alex Rosenblat, and danah boyd. 2015.
Discussion paper

Limitations of mitigating judicial bias with machine learning
Kristian Lum. 2017.

Trade-offs and impossibility results

Classification, Calibration, Precision, Recall

Probabilistic Outputs for Support Vector Machines and Comparisons to Regularized Likelihood Methods
John C. Platt. 1999.

Inherent Trade-Offs in the Fair Determination of Risk Scores
Jon Kleinberg, Sendhil Mullainathan, Manish Raghavan. 2016.

Fair prediction with disparate impact: A study of bias in recidivism prediction instruments
Alexandra Chouldechova. 2016.

Attacking discrimination with smarter machine learning
An interactive visualization by Martin Wattenberg, Fernanda Viégas, and Moritz Hardt. 2016.

Algorithmic decision making and the cost of fairness
Sam Corbett-Davies, Emma Pierson, Avi Feller, Sharad Goel, Aziz Huq. 2017.

The problem of Infra-marginality in Outcome Tests for Discrimination
Camelia Simoiu, Sam Corbett-Davies, Sharad Goel. 2017.

Inherent limitations of observational measures

Equality of Opportunity in Supervised Learning
Moritz Hardt, Eric Price, Nathan Srebro. 2016.

Beyond observational measures

Causal reasoning

Background reading:
Pearl (Chapter 1–3)
Pearl (Section 4.5.3)

Elements of Causal Inference
Peters, Janzing, Schölkopf

On causal interpretation of race in regressions adjusting for confounding and mediating variables
Tyler J. VanderWeele and Whitney R. Robinson. 2014.

Causal fairness criteria

Counterfactual Fairness
Matt J. Kusner, Joshua R. Loftus, Chris Russell, Ricardo Silva. 2017.

Avoiding Discrimination through Causal Reasoning
Niki Kilbertus, Mateo Rojas-Carulla, Giambattista Parascandolo, Moritz Hardt, Dominik Janzing, Bernhard Schölkopf. 2017.

Fair Inference on Outcomes
Razieh Nabi, Ilya Shpitser