Astraea: Grammar-based Fairness Testing

TL;DR

Astraea is a grammar-based testing tool that generates discriminatory inputs to identify and diagnose fairness violations in software, helping improve fairness in machine learning systems.

Contribution

Astraea introduces a novel grammar-based approach using probabilistic grammars for fairness testing and fault diagnosis in ML software systems.

Findings

Generated over 573,000 discriminatory test cases

Detected over 102,000 fairness violations

Improved software fairness by approximately 76%

Abstract

Software often produces biased outputs. In particular, machine learning (ML) based software are known to produce erroneous predictions when processing discriminatory inputs. Such unfair program behavior can be caused by societal bias. In the last few years, Amazon, Microsoft and Google have provided software services that produce unfair outputs, mostly due to societal bias (e.g. gender or race). In such events, developers are saddled with the task of conducting fairness testing. Fairness testing is challenging; developers are tasked with generating discriminatory inputs that reveal and explain biases. We propose a grammar-based fairness testing approach (called ASTRAEA) which leverages context-free grammars to generate discriminatory inputs that reveal fairness violations in software systems. Using probabilistic grammars, ASTRAEA also provides fault diagnosis by isolating the cause of observed software bias. ASTRAEA's diagnoses facilitate the improvement of ML fairness. ASTRAEA was evaluated on 18 software systems that provide three major natural language processing (NLP) services. In our evaluation, ASTRAEA generated fairness violations with a rate of ~18%. ASTRAEA generated over 573K discriminatory test cases and found over 102K fairness violations. Furthermore, ASTRAEA improves software fairness by ~76%, via model-retraining.