This talk will consider how we can continuously enhance and personalize any real interface using AI for adaptive A/B testing. The use of Bayesian multi-armed bandit algorithms to automatically conduct and improve A/B testing, as well as the necessary statistical techniques to analyze and draw valid inferences from this data will be discussed. Illustrations for using AI for social good by incorporating cognitive, social, and clinical psychology, in learning on mental and physical health will be provided. The application of these techniques to solving business problems, such as improving websites, emails, recommender systems and other specific user interfaces and products will also be discussed.
Joseph Jay Williams is an Assistant Professor in Computer Science at the University of Toronto, has courtesy appointments in Statistical Sciences & Psychology (to accept & supervise Ph.D. students), and is a Vector Institute for Artificial Intelligence Faculty Affiliate. He leads the Intelligent Adaptive Interventions research group of 11 graduate students across HCI, Psychology, applied ML, & Statistics.