In this 90 minutes workshop, we will explore bias in word embeddings - a widespread building block of many machine learning models that work with human languages. Word embeddings have an easy-to-explain representation that allows an intuitive understanding of this building block and its potential biases. They will serve as a case-study to the general issue of bias in machine learning.
Through the exploration, we will raise practical, methodological and philosophical questions about the ethics of AI. The workshop is designed to be adaptive to a diverse audience: from without any background in machine learning or programming to data science practitioners - no technical background is required! The workshop is hands-on and interactive, and will use the same tools that data scientists are using.
- Word embeddings will serve as a case-study to the general issue of bias in machine learning.
- The exploration process will naturally raise practical, methodological, and philosophical questions about the ethics of AI and the limitations of technical measurement and mitigation approaches.
- The materials are released under CC license, and if there is a demand, I'd be happy to produce an instructor guide for others.
- Being available in the Slack channel.
The workshop is built around hands-on activities that were delivered multiple times to a diverse audience. The materials (Jupyter Notebooks on the cloud) are designed to be easy to use (basically, clicking on a play bottom for each section), yet interactive. I feel comfortable with giving this workshop up to 60 participants.
Shlomi is a computer science Ph.D. student at Boston University working on responsible AI. In his previous life, Shlomi was a social entrepreneur and research team leader in cybersecurity.