Image: Generated by HM News with AI in Adobe Firefly
The results are in: Harvard Medical School Dean George Q. Daley has awarded funding to 33 projects through the inaugural Dean’s Innovation Awards for the Use of Artificial Intelligence in Education, Research, and Administration.
The funded projects highlight what is possible when humans leverage the rapidly evolving generative AI technology to advance the School’s mission of improving health and well-being for all.
“Through their innovative approaches to enhancing medical education, research, culture, and community, our inaugural award recipients’ projects exemplify the tremendous promise artificial intelligence holds at HMS and beyond,” said Daley.
First announced in autumn 2023, the awards were developed to stimulate creative ideas that use generative AI tools to:
- Transform medical and graduate education, including new technologies for interactive simulation-based learning, adaptive tutoring systems, and automated assessment.
- Open frontiers in medical and biological research across disciplines such as medical imaging, drug discovery, genomics, structural analysis, and public health informatics.
- Improve administrative efficiencies and productivity, such as through automation of workflows and improvements in data analytics and reporting.
Winning projects cover a range of applications, including generating images of diverse skin tones for use in dermatology teaching models, engineering proteins to deepen biological understanding and develop new drugs, improving patient diagnosis and treatment, helping faculty write and assess CVs, and strengthening social connectivity within the HMS community.
Each project will receive a one-year grant of up to $100,000 with the possibility to extend or apply for a second year.
Personalized teaching
Some of the projects will focus on using what are being called AI tutors in graduate and medical education.
Joseph Loparo, professor of biological chemistry and molecular pharmacology and director of educational programs in the Department of Biological Chemistry and Molecular Pharmacology in the Blavatnik Institute at HMS, is looking to push the boundaries of graduate student learning by using generative AI to revamp experimental design practice in a graduate-level molecular biology course and to create AI tutors to provide a personalized learning experience for all students.
According to Loparo, one challenge in education is offering students a personalized learning experience in which they are actively engaging with course materials.
“Historically, we have used structured small-group discussions in class to provide personalization,” Loparo said. “AI will allow us to reach the logical conclusion of this vision by offering opportunities for engagement on the individual student level.”
Sanjat Kanjilal, assistant professor of population medicine at Harvard Pilgrim Health Care Institute and director of the Mechanisms of Microbial Pathogenesis course at HMS, is co-leading a group in building an adaptive microbiology tutor that will use a Socratic approach that emulates real-life medical student-teacher interactions.
Each student has a unique set of experiences that can support or hinder their ability to learn new concepts, Kanjilal said. This can present a fundamental challenge for instructors who must ensure that every student finishes the course with a common understanding of core concepts. Kanjilal hopes the new project will support efforts to ensure educational equity.
“We hypothesize that the AI tutor will serve to reduce disparities in class performance across student demographics by providing personalized instruction with validated cases available 24/7,” he said.
He added that an AI tutor like this should allow teachers to spend more classroom time on challenging and nuanced topics.
Administrative efficiency
The approach to using AI to free up time for people to do what only they can do is also the basis of a project led by Melissa Korf, on using AI tools to improve the efficiency and quality of agreement reviews.
Korf, senior director of research contracts, data, and security at HMS, and her colleagues in the HMS Office of Research Administration support academic research activities through the review, negotiation, signing, and administration of research-related agreements.
With the new project, they aim to develop and pilot an implementation plan for an AI-based solution to increase the efficiency and quality of the agreement review process.
“We hope that this project will reduce the amount of time that ORA negotiators need to spend manually pulling information from various negotiation resources and enable them to spend more time using their expertise to determine the most effective negotiation strategy,” said Korf.
Advances in AI for research
Another winning project will look at overcoming challenges that current AI use can present.
For instance, protein language models, or PLMs, can glean protein structure based on the protein’s amino acid sequences. However, these models are not great at explaining protein function. Yet, understanding how proteins function and interact with each other is critical for the development of new therapies.
The new project involves developing PLMs that create more comprehensive models that are better at understanding and explaining protein function to enable the design of treatments. Project lead Marinka Zitnik, HMS assistant professor of biomedical informatics in the Blavatnik Institute, said this approach will bridge the gap between human and protein languages, allowing users to ask questions in natural language about molecular functions and the related physical characteristics that stem from their molecular structure.
“Our model will have several unique capabilities made possible using cutting-edge advances in generative AI that cannot be replicated using traditional deep-learning methods alone,” she said.
Zitnik added that the application of AI in research is a particularly promising endeavor. “I am excited by the rapid advances in multimodal text and image generation, generative AI, and multitask autonomous agents. I anticipate that these developments will soon help in the search for cures for diseases with no current treatments and the development of new, more effective drugs with fewer side effects, catering to specific patient subpopulations.”
