Penn awarded collaborative NSF grant to launch AI health institute

Penn has announced a partnership with the University of Delaware and several other collaborating universities to develop an institute focused on artificial intelligence’s role in health care.

Titled the National Science Foundation AI Institute for Human-AI Cooperation, the collaboration is funded by a $21.5 million, five-year award from the NSF and will bring together over 40 partners across academia, industry, and government. The institute is “designed to support, not replace, human decision making in healthcare,” according to a July 29 release.

HAIC director Cathy Wu said in an interview with The Daily Pennsylvanian that the institute seeks to “transform artificial intelligence into an adaptive teammate,” allowing humans to have “dynamic cooperation” with AI.

Wu explained that such collaboration would enable HAIC to develop “personalized models tailored to individual needs,” both creating “a new path for health care” and cultivating “the next generation of leaders in AI.”

Computer science professor and HAIC managing director Eric Eaton told the DP the institute aims to develop AI systems as “effective clinical partners” that do not “replace human clinical judgment.”

“What we’re really trying to build is a national hub for research on how humans and AI can work together effectively,” Eaton added.

According to HAIC’s website, the institute will interpret “multiscale and multimodal data” to help “make rehabilitation more personalized, effective, and responsive.”

In addition, HAIC will “use AI‑enabled digital twins to create virtual versions (or avatars) of patients, allowing clinicians to refine treatment options before applying them in real-world care,” according to the release.

SAS to launch Chief AI Officer Initiative, collaborate with local governments

Penn pilots secure AI portal with Claude, ChatGPT models

Eaton highlighted Penn’s GRASP Lab as a source of “strengths in foundational AI and machine learning and robotics” for HAIC’s work.

“We’re going to be pursuing research multimodal reasoning of time-varying processes, such as how a patient changes during treatment, adaptive and continual learning, where the system is persistent over a long period of time and adaptive with experience, and methods for human-AI interaction,” Eaton said.

He added that Penn Engineering and Penn Medicine are already collaborating on projects using multimodal clinical data “to study how AI can better support real-time clinical decision making.”

HAIC deputy director Austin Brockmeier told the DP that physical therapy is a “journey,” adding that the researchers want to establish “a triad between the AI, the patient, and the clinician.”

“We want the AI to provide a second opinion, doing things that a human can’t do,” Brockmeier said. “You can increase the amount of information that’s being measured and make that actionable for the treatment plan.”

Brockmeier explained that AI tools can also improve patients’ experiences outside the physician’s office, stating that “having additional measurements at home, as well as additional adaptive guidance during those exercises, can be transformative for them.”

Wu noted that the HAIC’s work is made possible by a “really exciting team effort” by the collaborating universities, each of which brings “all different kinds of expertise.”

She added that “the foundational AI technology and methodology” the researchers hope to develop could translate into “much broader economic and societal impact.”