Differential Tweetment - Mitigating Racial Dialect Bias in Harmful Tweet Detection
Association for Computing Machinery (ACM) via YouTube
Overview
Explore a conference talk that delves into the challenges of racial dialect bias in harmful tweet detection systems and proposes innovative solutions. Learn about the consequences of data set bias in social media content moderation and discover bias mitigation techniques such as adversary debiasing. Examine the performance results of these techniques and understand their key takeaways. Engage with thought-provoking questions surrounding predatory inclusion and consider future research directions in this critical area of natural language processing and social media ethics.
Syllabus
Introduction
Overview
Harmful Tweets
The Problem
The Consequences
Data Set Bias
Motivation
Bias Mitigation Techniques
Adversary Debiasing
Performance
Results
Key takeaways
Big questions
Predatory inclusion
Future research
Taught by
ACM FAccT Conference