Generalizing Orthogonalization for Models with Non-Linearities
This addresses bias risks in AI applications, such as preventing racial information from influencing medical decisions, though it builds incrementally on existing linear orthogonalization approaches.
The paper tackles the problem of biases in black-box algorithms by extending orthogonalization methods to handle non-linearities like ReLU activations, enabling protection of sensitive data in neural networks and other models. The method was validated through experiments on generalized linear models, convolutional neural networks, and pre-existing embeddings.
The complexity of black-box algorithms can lead to various challenges, including the introduction of biases. These biases present immediate risks in the algorithms' application. It was, for instance, shown that neural networks can deduce racial information solely from a patient's X-ray scan, a task beyond the capability of medical experts. If this fact is not known to the medical expert, automatic decision-making based on this algorithm could lead to prescribing a treatment (purely) based on racial information. While current methodologies allow for the "orthogonalization" or "normalization" of neural networks with respect to such information, existing approaches are grounded in linear models. Our paper advances the discourse by introducing corrections for non-linearities such as ReLU activations. Our approach also encompasses scalar and tensor-valued predictions, facilitating its integration into neural network architectures. Through extensive experiments, we validate our method's effectiveness in safeguarding sensitive data in generalized linear models, normalizing convolutional neural networks for metadata, and rectifying pre-existing embeddings for undesired attributes.