Despite widespread recognition of the importance of responsible machine learning, a gap between theory and practice is preventing the adoption of ethical guidelines. Our open-source best practices help bridge this gap by providing a comprehensive and pragmatic starting point for the implementation of responsible machine learning in both product development and organizational management.
The fast and often reckless adoption of machine learning(ML) in recent years has led to some solutions with undesirable consequences. Many examples have received plenty of media and public attention. This has spurred some good first steps directed at addressing these issues at an organizational and institutional level. There are some guides on theoretical principles by which an ethical ML solution should abide. However, there is little to no guidance on how to apply these principles in practice or on what it means to comply with them. The Foundation for Best Practices in Machine Learning aims to fill in that gap. Our motto is to ‘Champion ethical and responsible machine learning through open-source best practices and free public knowledge’. Our members are volunteers and have diverse backgrounds: data scientists, machine learning engineers, statisticians, legal professionals, academics, and communications experts. During this talk we plan to present our technical and organizational best practices. The technical best practices provide advice on how to develop and maintain a ML product responsibly, and include practical tips and tricks with respect to fairness and non-discrimination, explainability, monitoring and maintenance, security, etc. The organizational best practices provide advice on how to effectively support product teams across unique organizational contexts, touching upon managerial oversight, responsibility ownership, data governance, human resource and software management.