Text-to-image diffusion models often exhibit biases toward specific demographic groups, such as generating more males than females when prompted to generate images of engineers, raising ethical concerns and limiting their adoption. In this paper, we tackle the challenge of mitigating generation bias towards any target attribute value (e.g., "male" for "gender") in diffusion models while preserving generation quality. We propose FairGen, an adaptive latent guidance mechanism which controls the generation distribution during inference. In FairGen, a latent guidance module dynamically adjusts the diffusion process to enforce specific attributes, while a memory module tracks the generation statistics and steers latent guidance to align with the targeted fair distribution of the attribute values. Further, given the limitations of existing datasets in comprehensively assessing bias in diffusion models, we introduce a holistic bias evaluation benchmark HBE, covering diverse domains and incorporating complex prompts across various applications. Extensive evaluations on HBE and Stable Bias datasets demonstrate that FairGen outperforms existing bias mitigation approaches, achieving substantial bias reduction (e.g., 68.5% gender bias reduction on Stable Diffusion 2). Ablation studies highlight FairGen's ability to flexibly and precisely control generation distribution at any user-specified granularity, ensuring adaptive and targeted bias mitigation.

Recently, transductive learning methods, which leverage holdout sets during training, have gained popularity for their potential to improve speed, accuracy, and fairness in machine learning models. Despite this, the composition of the holdout set itself, particularly the balance of sensitive sub-groups, has been largely overlooked. Our experiments on CIFAR and CelebA datasets show that compositional changes in the holdout set can substantially influence fairness metrics. Imbalanced holdout sets exacerbate existing disparities, while balanced holdouts can mitigate issues introduced by imbalanced training data. These findings underline the necessity of constructing holdout sets that are both diverse and representative.

With the rising adoption of Machine Learning across the domains like banking, pharmaceutical, ed-tech, etc, it has become utmost important to adopt responsible AI methods to ensure models are not unfairly discriminating against any group. Given the lack of clean training data, generative adversarial techniques are preferred to generate synthetic data with several state-of-the-art architectures readily available across various domains from unstructured data such as text, images to structured datasets modelling fraud detection and many more. These techniques overcome several challenges such as class imbalance, limited training data, restricted access to data due to privacy issues. Existing work focusing on generating fair data either works for a certain GAN architecture or is very difficult to tune across the GANs. In this paper, we propose a pipeline to generate fairer synthetic data independent of the GAN architecture. The proposed paper utilizes a pre-processing algorithm to identify and remove bias inducing samples. In particular, we claim that while generating synthetic data most GANs amplify bias present in the training data but by removing these bias inducing samples, GANs essentially focuses more on real informative samples. Our experimental evaluation on two open-source datasets demonstrates how the proposed pipeline is generating fair data along with improved performance in some cases.

Last week Microsoft Corp said it would stop selling software that guesses a person's mood by looking at their face. The reason: It could be discriminatory. Computer vision software, which is used in self-driving cars and facial recognition, has long had issues with errors that come at the expense of women and people of color. Microsoft's decision to halt the system entirely is one way of dealing with the problem. But there's another, novel approach that tech firms are exploring: training AI on "synthetic" images to make it less biased.

Generative Adversarial Networks (GANs) have recently advanced face synthesis by learning the underlying distribution of observed data. However, it will lead to a biased image generation due to the imbalanced training data or the mode collapse issue. Prior work typically addresses the fairness of data generation by balancing the training data that correspond to the concerned attributes. In this work, we propose a simple yet effective method to improve the fairness of image generation for a pre-trained GAN model without retraining. We utilize the recent work of GAN interpretation to identify the directions in the latent space corresponding to the target attributes, and then manipulate a set of latent codes with balanced attribute distributions over output images. We learn a Gaussian Mixture Model (GMM) to fit a distribution of the latent code set, which supports the sampling of latent codes for producing images with a more fair attribute distribution. Experiments show that our method can substantially improve the fairness of image generation, outperforming potential baselines both quantitatively and qualitatively. The images generated from our method are further applied to reveal and quantify the biases in commercial face classifiers and face super-resolution model.