Presentation
Design Principles for Accessibility and Transparency to Facilitate Co-Learning in Human-AI Teams
DescriptionHuman-AI Teaming (HAT) holds exceptional promise in helping humans perform tasks that are typically difficult for humans to perform alone. With an ever-growing volume of multimedia data, intelligence analysts are faced with an increasingly insurmountable task; sifting through mountains of data to create the most complete analyses or reports. While there are AI tools to address this problem partially, these solutions present challenges for users that are not experts in AI applications. First, there is a significant cost to upkeep underlying models. Second, a lack of transparency and access to models creates usability issues for users who are not data scientists or software engineers. To address these challenges, we explore the concept of co-learning to improve HAT tool behavior with SCHOLA, an AI that continuously improves its ability to represent and reason about novel concepts, objects, and activities, so it and its human user learn together and from each other.
Contributors
Principal Engineer
Event Type
Lecture
TimeTuesday, September 10th3pm - 3:20pm MST
LocationFlagstaff
Human AI Robot Teaming (AI)