Course Summary
The course starts with introducing the fundamentals of algorithmic fairness, the most common biases in
machine learning, discusses recent examples from the algorithmic fairness in social media and gives an overview of several AI fairness tools meant to help engineers and data scientists examine, report, and mitigate discrimination and bias in ML models. This four-weeks course will include lectures and practical exercises with social media data that will help the audiences critically assess and mitigate the biases in data and in the algorithms too. During this course, we will present a comprehensive review of state-of-the-art techniques in fairness on Natural Language Processing and Graph Mining. This four-weeks master class experience is tailored for researchers and working professionals in areas including data mining, artificial intelligence, social science and beneficial to the social media domain.
Relevance
Since social media data differs from conventional data, it is imperative to study its unique characteristics. It is necessary for educators and researchers to move beyond traditional AI algorithms optimized for predictive performance and embed ethical and legal principles in their design, training, and deployment to ensure social good while still benefiting from the huge potential of the AI technology. Unfortunately, this massive communication is sometimes characterized by harm and abuse, such as hate speech, discrimination etc. Recent studies have revealed that many widely-applied NLP and graph mining models could suffer from potential discrimination and misuse. Given the wide usage of graph embedding methods and natural language processing to model social data, the course goal is to present different methods and studies that expose and mitigate social biases that manifest in structural properties in graph datasets or contextual Natural Language Processing methods.
This course aim to enable researchers and practitioners in data mining, artificial intelligence, social science, as well as related areas to better understand how AI performs on critical subgroups, spurring research, dialogue, and make progress toward "AI for Social Good" by moving academic approaches to trustworthy AI from the lab into real-world practical use.
