The risk of racial bias while tracking influenza-related content on social media using machine learning

Abstract

Objective Machine learning is used to understand and track influenza-related content on social media. Because these systems are used at scale, they have the potential to adversely impact the people they are built to help. In this study, we explore the biases of different machine learning methods for the specific task of detecting influenza-related content. We compare the performance of each model on tweets written in Standard American English (SAE) vs African American English (AAE). Materials and Methods Two influenza-related datasets are used to train 3 text classification models (support vector machine, convolutional neural network, bidirectional long short-term memory) with different feature sets. The datasets match real-world scenarios in which there is a large imbalance between SAE and AAE examples. The number of AAE examples for each class ranges from 2% to 5% in both datasets. We also evaluate each model's performance using a balanced dataset via undersampling. Results We find that all of the tested machine learning methods are biased on both datasets. The difference in false positive rates between SAE and AAE examples ranges from 0.01 to 0.35. The difference in the false negative rates ranges from 0.01 to 0.23. We also find that the neural network methods generally has more unfair results than the linear support vector machine on the chosen datasets. Conclusions The models that result in the most unfair predictions may vary from dataset to dataset. Practitioners should be aware of the potential harms related to applying machine learning to health-related social media data. At a minimum, we recommend evaluating fairness along with traditional evaluation metrics.

A case against social networks as the only means for recruitment in scientific studies with humans

“…Namely, the likelihood of segregation in networks, an effect known as filter bubble [32,34], and the ranking signals that govern how message contents become available to users in a network [3,21,29]. Other issues have been identified derived from online recruitment, especifically participant retention [28], age bias [17] and racial and economical bias [19,31,50]. Some studies have shown that community-based methods, including personal references, local advertisements and pamphlets, indeed recruit more racially diverse samples than other methods [41], and may ensure better chances of participant retention [2,7,44,46,49].…”

“…In this sense, differences in age have be found between participants recruited via social networks * E-mail: alvaropastor@uoc.edu and other methods [17]. As well as a risk for racial and economical bias has been reported when using social networks data [19,31,50]. Moreover, the literature from a diversity of disciplines, including machine learning and data mining [6,23,36], and marketing [20,27,48], sheds light on the various management layers that aim to optimise links between users and the flow of messages.…”

Misguided Artificial Intelligence: How Racial Bias is Built Into Clinical Models

“…Machine learning starts with raw data harvested from an ever-growing collection of data sources. Electronic health records, administrative health records, data warehouses, social media data, 6 as well as population health data are collected and stored with various entities. 7 If the raw data available for training and validation is biased, the analytical results will be biased.…”

Statistical quantification of confounding bias in machine learning models

“…As a characteristic example, models trained to predict intelligence [ 16 , 17 ] might provide a statistically significant predictive performance by picking up solely on age-related variance [ 18 , 19 ]. Moreover, various types of systematic sampling bias, as well as stochastic group differences in the training sample, can result in confounded models (e.g., racially biased machine learning models [ 6 , 20 , 21 ]).…”