Fairness Assessments
Class Coverage
- How can we assess algorithmic fairness?
- What metrics are commonly used to measure fairness?
- What can’t be captured in data?
Pre-Class Readings
Please read the following:
- Ethical Machine Learning in Health Care (Chen et al. 2021)
- Millions of black people affected by racial bias in health-care algorithms (Nature)
Participation Questions
Submit to Gradescope by 10AM PT on Thursday, November 3rd
- “Individual fairness” refers to the practice of treating similar individuals similarly. What fairness metrics can you use to measure individual fairness?
- “Group fairness” refers to the process of equalizing some output metric(s) (positive predictions, error rates, etc.) across demographic subgroups. What fairness metrics can you use to measure group fairness?
- Can you think of any types of fairness other than individual and group? How would you measure them?
This Week’s Slides
Assigned for Week 07
- Complete Replication Part 02: Model Development and Fairness Assessment
- Complete next week’s readings
- If you signed up to present on Model Cards for Model Reporting (Mitchell et al.) or Fairness Through Awareness (Dwork et al.) come prepared to present next week and submit your presentation on Gradescope by 10AM PT on Thursday, November 10th (note the different time!)
- Submit your answers to next week’s participation questions to Gradescope by 10AM PT on Thursday, November 10th
- Remember that Week 7’s class is rescheduled to 12:30PM PT on Thursday, November 10th!