Prediction algorithms are increasingly common in neuroimaging research and offer many opportunities for clinical care and personalized medicine. However, inadvertent outcomes can lead to predictions that are influenced by race, gender, sexuality, and ethnicity, or privileging one arbitrary group of users over others. Social bias can be introduced due to factors such as small dataset, sampling bias, and researcher introduced error  among many others. This panel is aimed at discussing the relevant factors and possible feasible solutions on ways to conceptualize, measure, and mitigate social bias in machine learning.

OSR 2022 | Panel #5: Social Bias in Machine Learning