Title:Software Engineering for Fairness: A Case Study with Hyperparameter Optimization

Abstract:We assert that it is the ethical duty of software engineers to strive to reduce software discrimination. This paper discusses how that might be done. This is an important topic since machine learning software is increasingly being used to make decisions that affect people's lives. Potentially, the application of that software will result in fairer decisions because (unlike humans) machine learning software is not biased. However, recent results show that the software within many data mining packages exhibits "group discrimination"; i.e. their decisions are inappropriately affected by "protected attributes"(e.g., race, gender, age, etc.).
There has been much prior work on validating the fairness of machine-learning models (by recognizing when such software discrimination exists). But after detection, comes mitigation. What steps can ethical software engineers take to reduce discrimination in the software they produce?
This paper shows that making \textit{fairness} as a goal during hyperparameter optimization can (a) preserve the predictive power of a model learned from a data miner while also (b) generates fairer results. To the best of our knowledge, this is the first application of hyperparameter optimization as a tool for software engineers to generate fairer software.