Fried Receives NSF CAREER Award

LTI assistant professor Daniel Fried has received the prestigious award for his work on interaction between humans and AI systems

Daniel Fried, an assistant professor in the Language Technologies Institute at Carnegie Mellon University, has received a National Science Foundation (NSF) Faculty Early Career Development Program (CAREER) award for his research on improving interaction and communication between people and AI systems.

One of the NSF’s most prestigious offerings, the CAREER award is given each year to early career faculty “who have the potential to serve as academic role models” and “lead advances in the mission of their department or organization,” according to the foundation.

Fried’s work targets the friction that sometimes arises between AI systems and their human users when the complex tasks carried out by the systems require nuanced correction.

“AI agents now help with very complex tasks, like writing entire software applications, analyzing data or navigating the web,” Fried said. “But communicating corrections to agents can require substantial time and effort from people.”

Fried’s work seeks to develop ways for users to provide feedback that’s robust enough to comprehensively instruct the agent on what it needs to correct, without being so burdensome for the user that they wish they had just completed the task themselves.

To gauge the success of an agent, Fried and his team measure not only how satisfied human testers are with an agent’s output, but also the amount of effort the users had to put in to get that result.

The upshot, Fried explained, is more time gained for humans to do the things they want to be doing.

“Society is now spending enormous numbers of person-hours interacting with AI agents,” he said. “I want our collective interaction effort to go farther so we can all spend more time doing other things.”

Fried noted his gratitude to the NSF, his colleagues and his students, singling out two of them. “Zora Wang’s work on inducing reusable skills for agents and Saujas Vaduguru’s work on pragmatic program synthesis and multimodal instruction were particularly helpful, ” he said.

Fried joins LP Morency and Noah Smith as recipients of NSF CAREER awards from the LTI.