Carnegie Mellon Joins NSF Effort to Build Trustworthy AI Assistants

Researchers from Carnegie Mellon University's School of Computer Science will contribute to a newly launched National Science Foundation (NSF) AI Research Institute to develop artificial intelligence assistants capable of trustworthy, sensitive and context-aware interactions with people. The assistants could potentially be used in mental and behavioral health fields, where trust and safety are of the utmost importance.

The AI Research Institute on Interaction for AI Assistants (ARIA) is led by Brown University and is one of five new NSF AI Institutes announced Tuesday as part of a $100 million investment in research. ARIA will combine research on human and machine cognition, with the goal of creating AI systems that can interpret a person's unique behavioral needs and provide helpful feedback in real time.

SCS researchers Reid Simmons and Henny Admoni, both faculty members in the Robotics Institute, will contribute to ARIA's efforts to develop AI assistants capable of interactively learning how to carry out complex physical and cognitive tasks — like preparing a meal — from people. Their work will emphasize real-world impact by helping AI systems better understand how to support individuals without taking control, promoting user autonomy and a greater sense of agency.

"This institute gives us a unique opportunity to explore the boundaries of how AI systems can truly collaborate with people, not just by performing tasks, but also by learning from and adapting to them in meaningful ways," Simmons said. "It will allow us to build on existing research underway at Carnegie Mellon and further the goal to develop AI assistants that enhance human capability while respecting autonomy and social context."

The CMU team will also collaborate with Brown faculty to investigate how AI assistants can learn and apply social norms in human interactions, ensuring these systems are respectful, culturally aware and trustworthy. ARIA will bring together experts from across the nation in fields spanning computer science and machine learning, cognitive and behavioral science, law, philosophy, and education to understand what form such systems should take and how they could be safely and responsibly deployed.

ARIA's work will complement research Simmons and Admoni have already begun as part of the NSF AI Institute for Collaborative Assistance and Responsive Interaction for Networked Groups (AI-CARING), which seeks to develop AI systems for caregiving environments.

CMU's work on foundational research and the practical applications of AI has featured heavily in NSF AI Research Institutes. The university houses the AI Institute for Societal Decision Making (AI-SDM), which explores how humans and AI can collaborate to make informed decisions in critical situations. In addition to AI-SDM and AI-CARING, CMU faculty participate in the AI Institute for Artificial and Natural Intelligence, the AI Institute for Future Edge Networks and Distributed Intelligence, and the USDA-NIFA AI Institute for Resilient Agriculture.