Fairness GAN

By Prasanna Sattigeri, Samuel C. Hoffman, Vijil Chenthamarakshan, Kush R. Varshney. 2018

98 citations recorded by the source corpus.

In this paper, we introduce the Fairness GAN, an approach for generating a dataset that is plausibly similar to a given multimedia dataset, but is more fair with respect to protected attributes in allocative decision making. We propose a novel auxiliary classifier GAN that strives for demographic parity or equality of opportunity and show empirical results on several datasets, including the CelebFaces Attributes (CelebA) dataset, the Quick, Draw!\ dataset, and a dataset of soccer player images and the offenses they were called for. The proposed formulation is well-suited to absorbing unlabeled data; we leverage this to augment the soccer dataset with the much larger CelebA dataset. The methodology tends to improve demographic parity and equality of opportunity while generating plausible images.

Subfields: Computer Vision and Pattern Recognition, Law, Generative Models, AI Ethics, Fairness & Society

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Law

In Generative Models

In AI Ethics, Fairness & Society

Citations in the corpus

Understanding and Evaluating Racial Biases in Image Captioning (2021, background): “Proposed solutions to dataset bias include new data collection approaches [38, 35], manual data cleanup [74, 75], synthetic data generation [58, 20, 60] – or, in extreme cases, even withdrawing the dataset after insurmountable biases have been identified [13].”
Uncovering and Mitigating Algorithmic Bias through Learned Latent Structure (2019, methodology): “For example, (Sattigeri et al. 2018) used a generative adversarial network (GAN) to output a reconstructed dataset similar to the input but more fair with respect to certain attributes.”
Algorithmic Fairness (2020, background): “There are several previous papers in the domain of fair image classification [25, 46, 114, 124].”
What We Can't Measure, We Can't Understand: Challenges to Demographic Data Procurement in the Pursuit of Fairness (2020, background): “, [28, 49, 71, 88])1, or to identify and mitigate algorithmic discrimination for computationally identified groups without the use of demographics at all (e.”
Towards Intersectionality in Machine Learning: Including More Identities, Handling Underrepresentation, and Performing Evaluation (2022, methodology): “…groups by drawing from machine learning techniques for imbalanced data such as generating synthetic examples of underrepresented groups [102, 105, 111]; and evaluate on more subgroups by generalizing existing fairness definitions, such as equal opportunity or demographic parity, through…”
Pros and Cons of GAN Evaluation Measures: New Developments (2021, background): “Some recent works ( e.g. Xu et al. (2018); Yu et al. (2020) have tried to mitigate the bias in GANs or use GANs to reduce bias in other ML algorithms ( e.g. Sattigeri et al. (2019); McDuﬀ et al. (2019)).”
Fair Attribute Classification through Latent Space De-biasing (2020, methodology): “Table 3: Comparison of the Attractive classifier trained using synthetic data from Fairness GAN [35] and the classifier trained using our pairaugmented synthetic data.”
Fair Generative Modeling via Weak Supervision (2019, unknown): “, 2018) and FairnessGAN (Sattigeri et al., 2019).”
Say No to the Discrimination: Learning Fair Graph Neural Networks with Limited Sensitive Attribute Information (2020, unknown): “They could reduce the bias by modifying the training data through correcting labels [22, 56], revising attributes of data [14, 23], generating non-discriminatory labeled data [41, 48, 49], and obtaining fair data representations [3, 9, 13, 29–31, 33, 42, 54].”
Reducing sentiment polarity for demographic attributes in word embeddings using adversarial learning (2020, background): “However, adversarial learning techniques have recently proven to be a powerful tool for reducing unintended bias in machine learning [2, 8, 23, 28].”
Generative Adversarial Networks (GANs) (2020, methodology): “[225] proposed an GANs-based approach to generate fair dataset w.”
Inclusive GAN: Improving Data and Minority Coverage in Generative Models (2020, background): “However, these advances have come at a cost: there could be potential biases in the learned model against underrepresented data subgroups [49,57,45,16,17].”
CheXclusion: Fairness gaps in deep chest X-ray classifiers (2020, methodology): “…the many methods that could be used to de-bias the classifier, including representation learning [2], adversarial learning [51], and GANs [40, 47], focusing instead on quantifying the TPR disparities and underdiagnosis rate distribution in SOTA deep learning models that might be trained…”
Fairness in Machine Learning: A Survey (2020, methodology): “[241] illustrate this by using a GAN with fairness constraints to produce “unbiased” image datasets and [278] have evidenced similar results for structured data.”
Solving a Class of Non-Convex Min-Max Games Using Iterative First Order Methods (2019, background): “Examples of problems that are formulated under this framework include generative adversarial networks (GANs) [51], reinforcement learning [11], adversarial learning [53], learning exponential families [10], fair statistical inference [17, 56, 52, 37], generative adversarial imitation learning [6, 27”
Rényi Fair Inference (2019, methodology): “…data to a transformed space in which the dependencies between the class label and the sensitive attributes are removed Edwards & Storkey (2015); Hardt et al. (2016); Xu et al. (2018); Sattigeri et al. (2018); Raff & Sylvester (2018); Madras et al. (2018); Zemel et al. (2013); Louizos et al. (2015).”