The Values Encoded in Machine Learning Research

During her doctorate, Abeba Birhane found herself surrounded by those immersed in the models they were building and the data sets they were using- but no one was asking questions about what was in those data sets.

As AI can inherit the biases of their human builders, and any bias present in the data that feeds them the end result, mirrors society, including biases. Failing to recognise these risks can cause real-world harm that can impact individuals and society as a whole.

Birhane's research focus is on investigating the data sets that are increasingly shaping our world - exposing their biases and holding the corporations that design and profit from them accountable.