A systematization of the literature showing that dataset size and group distance to the decision boundary, not the choice of DP algorithm, are likely the decisive factors in privacy-induced unfairness.
Exploring the Unfairness of DP-SGD Across Settings
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
abstract
End users and regulators require private and fair artificial intelligence models, but previous work suggests these objectives may be at odds. We use the CivilComments to evaluate the impact of applying the {\em de facto} standard approach to privacy, DP-SGD, across several fairness metrics. We evaluate three implementations of DP-SGD: for dimensionality reduction (PCA), linear classification (logistic regression), and robust deep learning (Group-DRO). We establish a negative, logarithmic correlation between privacy and fairness in the case of linear classification and robust deep learning. DP-SGD had no significant impact on fairness for PCA, but upon inspection, also did not seem to lead to private representations.
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
SoK: What Makes Private Learning Unfair?
A systematization of the literature showing that dataset size and group distance to the decision boundary, not the choice of DP algorithm, are likely the decisive factors in privacy-induced unfairness.