Representation Bias in Data: A Survey on Identification and Resolution Techniques

Abstract

Data-driven algorithms are only as good as the data they work with, while data sets, especially social data, often fail to represent minorities adequately. Representation Bias in data can happen due to various reasons ranging from historical discrimination to selection and sampling biases in the data acquisition and preparation methods. Given that "bias in, bias out", one cannot expect AI-based solutions to have equitable outcomes for societal applications, without addressing issues such as representation bias. While there has been extensive study of fairness in machine learning models, including several review papers, bias in the data has been less studied. This paper reviews the literature on identifying and resolving representation bias as a feature of a data set, independent of how consumed later. The scope of this survey is bounded to structured (tabular) and unstructured (e.g., image, text, graph) data. It presents taxonomies to categorize the studied techniques based on multiple design dimensions and provides a side-by-side comparison of their properties. There is still a long way to fully address representation bias issues in data. The authors hope that this survey motivates researchers to approach these challenges in the future by observing existing work within their respective domains.

arXiv commentJust Accepted ACM Comput. Surv. (March 2023)

References in corpus (25)

• SMOTE: Synthetic Minority Over-sampling Technique
• Mixing patterns in networks
• Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings
• Inherent Trade-Offs in the Fair Determination of Risk Scores
• Bias in Bios: A Case Study of Semantic Representation Bias in a High-Stakes Setting
• Getting Gender Right in Neural Machine Translation
• Why Is My Classifier Discriminatory?
• FairDrop: Biased Edge Dropout for Enhancing Fairness in Graph Representation Learning
• Assessing and Remedying Coverage for a Given Dataset
• Stereotypical Bias Removal for Hate Speech Detection Task using Knowledge-based Generalizations
• Gender Bias in Coreference Resolution: Evaluation and Debiasing Methods
• Quantifying Bias in Automatic Speech Recognition
• Model Patching: Closing the Subgroup Performance Gap with Data Augmentation
• Conditional Network Embeddings
• Big Holes in Big Data: A Monte Carlo Algorithm for Detecting Large Hyper-rectangles in High Dimensional Data
• Auditing ImageNet: Towards a Model-driven Framework for Annotating Demographic Attributes of Large-Scale Image Datasets
• Active Sampling for Min-Max Fairness
• Adaptive Sampling for Minimax Fair Classification
• Black is to Criminal as Caucasian is to Police: Detecting and Removing Multiclass Bias in Word Embeddings
• Interpretable Data-Based Explanations for Fairness Debugging
• Toward Understanding Bias Correlations for Mitigation in NLP
• A Survey on Bias in Visual Datasets
• Identification of Bias Against People with Disabilities in Sentiment Analysis and Toxicity Detection Models
• Achieving Representative Data via Convex Hull Feasibility Sampling Algorithms
• Handling Bias in Toxic Speech Detection: A Survey

Cited by in corpus (7)

• Data Readiness for AI: A 360-Degree Survey
• Whither Bias Goes, I Will Go: An Integrative, Systematic Review of Algorithmic Bias Mitigation
• Bridging the gap: Towards an Expanded Toolkit for AI-driven Decision-Making in the Public Sector
• Representation Debiasing of Generated Data Involving Domain Experts
• Explanatory Debiasing: Involving Domain Experts in the Data Generation Process to Mitigate Representation Bias in AI Systems
• Survey on AI Ethics: A Socio-technical Perspective
• Mining the Minoria: Unknown, Under-represented, and Under-performing Minority Groups