Case Study Identifying and Mitigating Unintended Demographic Bias in Machine Learning for NLP
[MUSIC PLAYING] AUDACE NAKESHIMANA: In our work on fairness and AI, we present a case study on natural language processing titled "Identifying and Mitigating Unintended Demographic Bias in Machine Learning." We will break down what each part of the title means. This is the work that was done jointly by Chris Sweeney and Maryam Najafian. My name is Audace Nakeshimana. I am a Researcher at MIT, and I'll be presenting their work. The content of the slides presents a high-level overview of a thesis project that was done throughout a course of the year. It will be released soon on MIT DSpace. AI has the power to impact society in a vast amount of ways. For example, in the banking industry, many companies are trying to use machine learning to figure out if someone will default on a loan given the data about them. Now, because machine learning is used in the high-stakes applications, errors that cause it to be unfair could cause discrimination, preventing certain d...