Combating AI Bias at Scale
As the technical lead on Jigsaw’s Conversation AI team, Lucy Vasserman is an expert on natural language and speech processing. Her research focuses on how computers can learn to understand the nuances and context of abusive language at scale, with an emphasis on combating algorithmic bias. The Conversation AI team applies this research to develop tools that address toxicity and reduce harassment. Most notably, Jigsaw has created an API called Perspective that uses machine learning to help make conversations online less toxic. This talk will explore Jigsaw’s work to understand and reduce unintended bias in machine learning models. Lucy will discuss her team’s data gathering and annotation strategies as well as how the team has implemented new strategies for mitigating model bias to improve the effectiveness of their tools.