Fairness Notion
Fairness notions in machine learning aim to mitigate biases in algorithms that disproportionately affect certain groups, focusing on ensuring equitable outcomes across diverse populations. Current research explores various fairness definitions (e.g., statistical parity, equalized odds, counterfactual fairness), developing and comparing methods like adversarial debiasing, data augmentation, and post-processing techniques across different model architectures (including deep learning and reinforcement learning). This work is crucial for building trustworthy and ethical AI systems, impacting fields like healthcare, finance, and criminal justice by promoting fairness and reducing discriminatory outcomes.
Papers
On the Vulnerability of Fairness Constrained Learning to Malicious Noise
Avrim Blum, Princewill Okoroafor, Aadirupa Saha, Kevin Stangl
Towards Better Fairness-Utility Trade-off: A Comprehensive Measurement-Based Reinforcement Learning Framework
Simiao Zhang, Jitao Bai, Menghong Guan, Yihao Huang, Yueling Zhang, Jun Sun, Geguang Pu