Fairness assessment for deep generative models

A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by one or more processors, the program instructions executable to:

• receive a user defined fairness criterion for a deep generative model;
• probe the deep generative model to produce samples for a target output;
• evaluate the samples for a fairness of the deep generative model, according to the user defined fairness criterion, wherein an application programming interface provides access to a classifier that predicts distributional characteristics for a protected feature, wherein the classifier is trained using samples that are pre-labelled with the protected feature;
• in response to determining that the deep generative model does not meet the user defined fairness criterion, produce a set of recommendations for modifying the deep generative model to meet the user defined fairness criterion;
• in response to determining that the deep generative model is to be modified, apply at least one of the set of recommendations to the deep generative model;
• and
• update the deep generative model.

A computer system comprising one or more processors, one or more computer readable tangible storage devices, and program instructions stored on at least one of the one or more computer readable tangible storage devices for execution by at least one of the one or more processors, the program instructions executable to:

• receive a user defined fairness criterion for a deep generative model;
• probe the deep generative model to produce samples for a target output;
• evaluate the samples for a fairness of the deep generative model, according to the user defined fairness criterion, wherein an application programming interface provides access to a classifier that predicts distributional characteristics for a protected feature, wherein the classifier is trained using samples that are pre-labelled with the protected feature;
• in response to determining that the deep generative model does not meet the user defined fairness criterion, produce a set of recommendations for modifying the deep generative model to meet the user defined fairness criterion;
• in response to determining that the deep generative model is to be modified, apply at least one of the set of recommendations to the deep generative model;
• and
• update the deep generative model.