Procedural Justice in Algorithmic Fairness: Transparency, Control and Participation Perspectives
summary
Argues that procedural fairness—transparency, control and participation—matters as much as outcome metrics; compares a fair-division algorithm (used via Split It for chore allocation) to group discussion and finds 40% deem algorithmic assignments unfair while discussion rates higher. Builds a participatory pipeline (RebuildAI) with pairwise-trained individual models and Borda voting for a real deployment at 412 Food Rescue; individual models reach ~88% fit, simulations show more equal distributions versus historical human dispatch (20% of recipients previously got 70% of donations), and editable explanations raise perceived fairness.
AI-written summary · about
chapters (15)
- 0:00Introduction to study
- 2:03Algorithmic management
- 4:40Perceptions of fairness
- 6:40Procedural justice
- 9:07Fair division algorithms
- 11:23Pilot study challenges
- 16:37Task allocation experiment
- 17:51Discussion vs algorithm
- 21:47Procedural justice factors
- 24:51The 'We Build AI' framework
- 27:34Nonprofit case study
- 30:23Building stakeholder models
- 32:51Case study outcomes
- 39:58Giving control to users
- 42:35Future research directions
related talks
34mInterpretability and Algorithmic Fairness
35mOpen Source Fairness
42mFairness in Machine Learning: From Theory to Practice
21mFairness and Accountability Design Needs for Algorithmic Support in High-Stakes Public Sector ...
40mQuestions and Practices to Make Algorithmic Decision Making More Fair
32mOperationalizing Responsible AI
54mKeynote - Fairness and Diversity in Online Social Systems
11mAddressing Privacy, Bias, and Fairness in AI Systems