A More Equal Future?: Political Equality, Discrimination, and Machine Learning

Machine learning is everywhere. On social media platforms and news sites, in hiring, advertising, mortgage lending, criminal justice, education, and countless other sectors, more and more decisions are being made using predictions generated by algorithms that use complex data processing techniques. AI-evangelists promise that data-driven decision-making will not only boost organizational efficiency, but will also help make organizations fairer and advance social justice. By reducing the scope for human prejudice, irrationality, and error, they claim, machine learning can ensure decisions are made with complete consistency, treating each and every person without regard to morally irrelevant differences. Yet the effects of machine learning on social justice, human rights, and democracy will depend not on the technology itself, but on human choices about how to design and deploy it. Building and integrating machine learning models into decision-making systems involves choices that prioritize among the interests of different social groups and bake in different fundamental values. Among the most important is whether systems reproduce and entrench pervasive patterns of inequality and how to ensure they do not. How organizations respond to that issue will shape the implications of machine learning for equality, liberty, and fairness, the foundational principles of a flourishing constitutional democracy.