Algorithmic Fairness Trade-Offs in Sequential Selection
A new study on arXiv (2605.06227) examines the Price of Fairness (PoF) in algorithmic decision-making, focusing on sequential selection where decisions affect both immediate utility and future population distributions. The authors introduce short-term and long-term group fairness notions, showing that static fairness constraints can worsen long-run disparities. They find that short-term PoF can be large even when groups are nearly identical, but long-term disparities can vanish under simple investment policies with low PoF. The work is theoretical with empirical validation.
Key facts
- Study on arXiv (2605.06227) examines Price of Fairness in sequential selection
- Focuses on trade-off between fairness and utility in algorithmic decisions
- Introduces short-term and long-term group fairness notions
- Static fairness constraints may exacerbate long-run disparities
- Short-term PoF can be large even with nearly identical group distributions
- Long-term disparities can vanish under simple investment policies
- Theoretical analysis with empirical validation
Entities
Institutions
- arXiv