Harrison     
Brent Harrison
Research Scientist
Entertainment Intelligence Lab
Georgia Institute of Technology

"Using Human Data to Improve AI-Human Interaction"

Monday, April 4, 4:00 PM
Packard Lab room 466

Abstract:  In recent years, artificial intelligence (AI) has become more prevalent in our everyday lives. Presently, it is not unusual for AI systems (such as Siri and Cortana) and humans to interface with each other on a daily basis. This phenomenon will continue as technologies such as self-driving vehicles and robots for healthcare integrate themselves into the general public. Given this newfound emphasis on human interaction, it is imperative that modern AI systems be imbued with an improved understanding of human behavior, decision making, preferences, etc. Doing so allows these systems to better communicate and interact with humans.

In this talk, Brent will discuss his general research approach to improving AI systems by imbuing them with a greater knowledge of human decision making, behavior, and preferences in order to facilitate positive interactions between humans and these systems. He will focus on two projects that show how data describing humans can be used to improve AI-human interaction. The first project involves using human replay data in video games to create a system that dynamically adapts virtual environments in order to reduce churn. Brent will also discuss his more recent work on the Quixote system, which uses the procedural knowledge contained in stories to train virtual agents to exhibit believable behavior. In addition, Brent will discuss the future of this work and how it will further improve the complex relationship between humans and AI systems.

Bio:  Brent Harrison received the B.S. degree in Computer Science and the B.A. degree in English from Auburn University in 2008, and the M.S. degree in computer science from North Carolina State University in 2012. He received the Ph.D. degree in computer science from North Carolina State University in 2014 for his work on creating adaptive virtual environments using data-driven models of player behavior.

He is currently a Research Scientist in the Entertainment Intelligence Lab at the Georgia Institute of Technology where he studies how the knowledge contained within stories can be used to train virtual agents. His primary research interests involve developing better interactive AI systems and machine learning algorithms by leveraging different types of human data.

© 2014-2016 Computer Science and Engineering, P.C. Rossin College of Engineering & Applied Science, Lehigh University, Bethlehem PA 18015.