Fairness Drift: Assessing Algorithmic Fairness in a Moving Market
Updated
Updated · arxiv.org · Nov 9
Fairness Drift: Assessing Algorithmic Fairness in a Moving Market
1 articles · Updated · arxiv.org · Nov 9
Summary
Cook County’s adoption of a machine-learning property assessment system in Chicago halved the over-assessment of lower-priced homes, improving fairness in tax bills.
The reform replaced manual adjustments with a gradient-boosted model and richer data, but similar deployments in suburban areas showed little effect due to prior market shifts.
Researchers warn that algorithmic fairness can drift as markets move, suggesting fairness audits should be triggered by market changes, not just fixed schedules.