Identifying Bias in AI using Simulation
arXiv:1810.00471
Abstract
Machine learned models exhibit bias, often because the datasets used to train them are biased. This presents a serious problem for the deployment of such technology, as the resulting models might perform poorly on populations that are minorities within the training set and ultimately present higher risks to them. We propose to use high-fidelity computer simulations to interrogate and diagnose biases within ML classifiers. We present a framework that leverages Bayesian parameter search to efficiently characterize the high dimensional feature space and more quickly identify weakness in performance. We apply our approach to an example domain, face detection, and show that it can be used to help identify demographic biases in commercial face application programming interfaces (APIs).
References in corpus (6)
- Practical Bayesian Optimization of Machine Learning Algorithms
- Semantics derived automatically from language corpora contain human-like biases
- A Tutorial on Bayesian Optimization of Expensive Cost Functions, with Application to Active User Modeling and Hierarchical Reinforcement Learning
- The Dataset Nutrition Label: A Framework To Drive Higher Data Quality Standards
- From Facial Parts Responses to Face Detection: A Deep Learning Approach
- InclusiveFaceNet: Improving Face Attribute Detection with Race and Gender Diversity
Cited by in corpus (6)
- Characterizing Bias in Classifiers using Generative Models
- Advances in Deep Space Exploration via Simulators & Deep Learning
- Advancing Non-Contact Vital Sign Measurement using Synthetic Avatars
- Multiple-Identity Image Attacks Against Face-based Identity Verification
- Detecting discriminatory risk through data annotation based on Bayesian inferences
- Does the dataset meet your expectations? Explaining sample representation in image data