The field of artificial intelligence is moving towards developing more trustworthy and fair systems, with a focus on mitigating bias and ensuring fairness in various applications. Recent research has highlighted the importance of addressing fairness issues in AI, particularly in areas such as healthcare and finance, where biased datasets can have significant consequences.

Researchers are exploring innovative approaches to fairness interventions, including controllable feature whitening, imbalance mitigating entity augmentation, and auditing vulnerabilities to distributional manipulation attacks. These methods aim to improve the reliability and fairness of AI systems, while also providing insights into the trade-offs between utility and fairness.

Noteworthy papers in this area include:

• Controllable Feature Whitening for Hyperparameter-Free Bias Mitigation, which proposes a simple yet effective framework for mitigating bias in deep neural networks.
• Exposing the Illusion of Fairness: Auditing Vulnerabilities to Distributional Manipulation Attacks, which investigates methods for manipulating data samples to artificially satisfy fairness criteria and offers recommendations for detecting such manipulations.
• Improving Group Fairness in Tensor Completion via Imbalance Mitigating Entity Augmentation, which proposes a method for improving group fairness in tensor decomposition and demonstrates its effectiveness on various datasets.