|
Abstract:
|
What constitutes a fair or unfair algorithm is context specific. Metrics for evaluating fairness have been developed, as have methods for prioritizing measures of fairness when building algorithms. However, algorithms are not neutral and optimization choices will reflect a specific value system and the distribution of power to make these decisions. Data also reflect societal bias, such as structural racism. Algorithmic fairness research spans many fields, including sociology, ethics, and computer science, with fewer contributions appearing in the statistics literature. These concepts are not routinely incorporated in statistics research and teaching, despite their importance and the potential for real harm to marginalized groups. I will provide an introductory overview of algorithmic fairness for a statistics audience and discuss an ethical pipeline for algorithms.
|
