Adversarial Removal of Gender from Deep Image Representations

By Tianlu Wang, Jieyu Zhao, Mark Yatskar, Kai-Wei Chang, Vicente Ordonez. 2018

18 citations recorded by the source corpus.

In this work we analyze visual recognition tasks such as object and action recognition, and demonstrate the extent to which these tasks are correlated with features corresponding to a protected variable such as gender. We introduce the concept of natural leakage to measure the intrinsic reliance of a task on a protected variable. We further show that machine learning models of visual recognition trained for these tasks tend to exacerbate the reliance on gender features. To address this, we use adversarial training to remove unwanted features corresponding to protected variables from intermediate representations in a deep neural network. Experiments on two datasets: the COCO dataset (objects), and the imSitu dataset (actions), show reductions in the extent to which models rely on gender features while maintaining most of the accuracy of the original models. These results even surpass a strong baseline that blurs or removes people from images using ground-truth annotations. Moreover, we provide convincing interpretable visual evidence through an autoencoder-augmented model showing that this approach is performing semantically meaningful removal of gender features, and thus can also be used to remove gender attributes directly from images.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, 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 Artificial Intelligence

In Computer Vision and Pattern Recognition

In AI Ethics, Fairness & Society

Citations in the corpus

Don’t Take the Easy Way Out: Ensemble Based Methods for Avoiding Known Dataset Biases (2019, background): “A popular approach is to use an adversary to remove information about a target feature, often gender or ethnicity, from a model’s internal representations (Edwards and Storkey, 2016; Wang et al., 2018; Kim et al., 2019).”
Demoting Racial Bias in Hate Speech Detection (2020, methodology): “Adversarial training has been used to remove protected features from word embeddings (Xie et al., 2017; Zhang et al., 2018) and intermediate representations for both texts (Elazar and Goldberg, 2018; Zhang et al., 2018) and images (Edwards and Storkey, 2015; Wang et al., 2018).”
Automatically Neutralizing Subjective Bias in Text (2019, background): “For instance, Bordia and Bowman (2019) introduced a regularization term for the language model to penalize the projection of the word embeddings onto that gender subspace, while Wang et al. (2018)”
Hurtful words: quantifying biases in clinical contextual word embeddings (2020, methodology): “Finally, we attempt to correct for the baseline clinical BERT’s performance disparities using adversarial debiasing during pretraining [8, 42, 66, 71], where a discriminator model forces learned embeddings to be minimally predictive of protected subgroup information.”
Why Can't I Dance in the Mall? Learning to Mitigate Scene Bias in Action Recognition (2019, background): “Building upon adversarial training, recent work aims to mitigate unintended biases by using the adversary to predict the protected variables [68, 58] or quantify the statistical dependency between the inputs and the protected variables [1].”
Ethical Considerations in AI-Based Recruitment (2019, background): “However, as AI becomes more commonplace, there are several areas, such as image, audio, and video classification, where bias has been studied [10, 30], and methods for remedying such bias will need to be applied.”
Think Locally, Act Globally: Federated Learning with Local and Global Representations (2020, background): “A recent focus has been on learning fair representations [82], including using adversarial training [30] to learn representations that are not informative of predeﬁned private attributes [25, 16, 50, 54, 60] such as demographics [24] and gender [78].”