Academics / Courses / DescriptionsCOMP_SCI 396, 496: HCI With & For AI
VIEW ALL COURSE TIMES AND SESSIONS
Prerequisites
(Comp_sci 329 or Comp_sci 330) and (Comp_sci 348 or Comp_sci 349) or CS MS or CS PhDs or Instructor permissionDescription
Human interaction with AI systems, and the use of AI to enable human interaction with computer systems (whether AI or not) creates amazing opportunities along with significant challenges. In the first case, how can users gain the value offered by AI while understanding the limits of AI systems and navigating the boundaries of their competence? How can they know when these systems are working effectively, and when they aren’t — when to trust them, and when not to? What kinds of interfaces and interactions will enable users to appropriately task these systems, understand and appropriately use their output or otherwise deploy them (including reasonably assessing the uncertainty associated with outputs), detect and correct errors and biases, and steer them appropriately towards better results? How do different kinds of tasks (ideation, critiquing, editing / debugging), different modalities (linguistic, visual, etc.) and different domains affect the nature of the interaction that must be supported to gain value while avoiding pitfalls?
In the second case, how can the use of AI for user interaction enable more fluent and natural interaction, while at the same time appropriately conveying the scope and abilities of the system(s) with which the user is interacting? How can such systems be designed to be more robust and reliable, and how can those properties be assessed? How can they be scoped, how can appropriate data be obtained, how can they be stress-tested? How do we cope with edge cases that may not be well-represented in available data distributions? What aspects of human communication and interaction transcend specific applications and must be understood by system designers and managed by interactive AI systems to achieve fluent and robust performance.
These questions are vast: This course aims to make a start at understanding the larger context of these questions, the current state of the art, and some specific lessons we can learn and incorporate into practice. The course will include both a reading seminar component and a projects component. We will read and discuss key papers addressing the topics described above; we will also build interactive AI projects aimed at incorporating lessons learned and extending the state of the art. Robust conversational interaction will be a key theme of both our readings / discussions, and our project work.
The course is intended for advanced undergraduate students (juniors and seniors) and graduate students (MS and PhD). Students must have taken CS 348 or CS 349 (preferably both); and CS 330 or CS 329.
- This course fulfills the Technical Elective area.
REQUIRED TEXTBOOK: N/A
COURSE COORDINATOR: Prof. Larry Birnbaum
COURSE INSTRUCTOR: Prof. Birnbaum