Algorithmic bias, and potential negative outcomes of machine learning, have gained deserved attention. However, relatively few standard processes exist for industry practitioners to assess and address algorithmic bias and unintended outcomes. This lecture will discuss challenges encountered in practice and at scale, and adds domain-specific examples from the music domain. We describe the 'translation' of existing literature frameworks; and practical efforts in understanding data, modeling and measurement decisions' potential impact. We'll share technical and organizational lessons learned, data challenges and interpretation pitfalls.
Lecture Slides will appear here soon.
Video footage will be added in the future.