Ray Hong is a researcher who specialized in Human-Computer Interaction.
His emphasis is to help human users to better leverage Deep Neural Networks (DNNs) in the context below:
1. Human-AI Collaboration: My research group design interactive systems that leverage DNN-driven automation to support professional workers. In particular, we are developing interactive tools for radiologists who simplify their professionally written sentences to improve readability for a broader audience. We also design interactive tools for comic artists who can significantly reduce their workload in colorizing the voluminous amount of images for rolling out their weekly stories.
2. Human-in-the-loop: I enable professional data scientists to indicate their thoughts directly and intuitively through interactive design. We call this design Interactive Attention Alignment or Explanation-based Supervision. In particular, we help data scientists to spot the vulnerability of DNNs through a local level of explanation, help them intelligently change the boundary of explanation based on their perspectives, and update DNN’s future model behavior. With our design, we contribute to making more fair and accountable DNNs.
3. Interactive Data Annotation: The quality of DNNs is the quality of data. We aim at designing interactive data annotation user interfaces used by human users. We also explore the ways how we can incorporate marginalized/vulnerable people in the workspace of data annotation.