Learning Fairer Representations with FairVIC
Keywords: Bias Mitigation, Fairness, Deep Learning
TL;DR: FairVIC enhances fairness in deep learning by integrating variance, invariance, and covariance terms into the loss function, reducing bias without compromising accuracy and achieving strong performance across multiple fairness metrics.
Abstract: Mitigating bias in automated decision-making systems, specifically deep learning models, is a critical challenge in achieving fairness. This complexity stems from factors such as nuanced definitions of fairness, unique biases in each dataset, and the trade-off between fairness and model accuracy. To address such issues, we introduce FairVIC, an innovative approach designed to enhance fairness in neural networks by addressing inherent biases at the training stage. Unlike other methods that require a user-defined declaration of what it means to be fair, FairVIC integrates an abstract concept of fairness through variance, invariance and covariance terms into the loss function. These terms aim to minimise the model's dependency on protected characteristics for making predictions, thus promoting fairness. Our experimentation consists of evaluating FairVIC against other comparable bias mitigation techniques, on a number of datasets known for their biases. Additionally, we conduct an ablation study to examine the accuracy-fairness trade-off. We also extend FairVIC by offering multi-objective lambda recommendations, allowing users to train a fairer model with a set of weights that are tuned best for their application. Through our implementation of FairVIC, we observed a significant improvement in fairness across all metrics tested, without compromising the model's accuracy. Our findings suggest that FairVIC presents a straightforward, out-of-the-box solution for the development of fairer deep learning models, thereby offering a generalisable solution applicable across many tasks and datasets.
Primary Area: alignment, fairness, safety, privacy, and societal considerations
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 6421
Loading