The role of tech companies in governing and designing ethical systems

Machine learning and artificial intelligence are used across numerous Facebook products. Facebook has been actively developing processes for the measurement of its machine learning and AI systems to ensure that they are not biased against users. This presentation gives an overview of algorithmic fairness, including an explanation of the ways that large-scale systems can be vulnerable to biases and a description of multiple, incompatible approaches of measuring fairness. It also provides insights into ways that Facebook evaluates models for possible biases and scales fairness tooling and frameworks across products, highlighting processes and insights for ensuring fairness across large scale systems that impact billions of users.