Biases in AI and How to tackle them

Bias is a complex societal notion that has gained traction among Artificial Intelligence researchers and practitioners. As we use AI-enabled systems today, from unlocking our phones to deciding one's creditworthiness, it has become crucial to define what are the biases and harms of Machine Learning Systems and find ways to mitigate them. There are many definitions for Algorithmic Bias, the one I choose today is of Kate Crawford 2017, "A skew that produces a type of harm". In 2018, Reuters reported that Amazon secretly scrapped their resume screening tools as it was being unfair to women candidates. Engineers at Amazon trained a machine learning model to score candidates profiles based on their resumes.