Stereotype-Free Classification of Fictitious Faces

TL;DR

This paper introduces a statistical method using penalized regression to classify GAN-generated faces without human bias, promoting fairness in AI face recognition.

Contribution

It proposes a novel penalized regression approach to label synthetic faces, reducing stereotype bias in AI classification tasks.

Findings

Effective in classifying GAN faces without human bias

Reduces stereotype influence in AI face recognition

Provides a new statistical tool for fair AI classification

Abstract

Equal Opportunity and Fairness are receiving increasing attention in artificial intelligence. Stereotyping is another source of discrimination, which yet has been unstudied in literature. GAN-made faces would be exposed to such discrimination, if they are classified by human perception. It is possible to eliminate the human impact on fictitious faces classification task by the use of statistical approaches. We present a novel approach through penalized regression to label stereotype-free GAN-generated synthetic unlabeled images. The proposed approach aids labeling new data (fictitious output images) by minimizing a penalized version of the least squares cost function between realistic pictures and target pictures.