English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Current adoption of machine learning in industrial, societal and economicalactivities has raised concerns about the fairness, equity and ethics ofautomated decisions. Predictive models are often developed using biaseddatasets and thus retain or even exacerbate biases in their decisions andrecommendations. Removing the sensitive covariates, such as gender or race, isinsufficient to remedy this issue since the biases may be retained due to otherrelated covariates. We present a regularization approach to this problem thattrades off predictive accuracy of the learned models (with respect to biasedlabels) for the fairness in terms of statistical parity, i.e. independence ofthe decisions from the sensitive covariates. In particular, we consider ageneral framework of regularized empirical risk minimization over reproducingkernel Hilbert spaces and impose an additional regularizer of dependencebetween predictors and sensitive covariates using kernel-based measures ofdependence, namely the Hilbert-Schmidt Independence Criterion (HSIC) and itsnormalized version. This approach leads to a closed-form solution in the caseof squared loss, i.e. ridge regression. Moreover, we show that the dependenceregularizer has an interpretation as modifying the corresponding Gaussianprocess (GP) prior. As a consequence, a GP model with a prior that encouragesfairness to sensitive variables can be derived, allowing principledhyperparameter selection and studying of the relative relevance of covariatesunder fairness constraints. Experimental results in synthetic examples and inreal problems of income and crime prediction illustrate the potential of theapproach to improve fairness of automated decisions.

Kernel Dependence Regularizers and Gaussian Processes with Applications to Algorithmic Fairness