Starting the Conversation: Implications of Generative AI for Gerontology

In her opening remarks, Dean Antonia Villarruel said, “Penn Nursing has long been a leader in the care of older adults – both in science and practice. Consistent with the mission and values of our School, we have a strong focus on those most vulnerable. We are elevating our current work by building strength in Artificial Intelligence to address opportunities in research and to see how we can apply this technology to issues confronted by individuals and families in health care settings and in homes and communities. I am confident your time together will generate an important path forward.”

In the space of gerontology, the use of Generative AI is anticipated to inform numerous applications including supporting and automating clinical decision making, and even predicting cognitive decline based on end user data that include audio and text, providing digital companionship chatbots for isolated older adults, processing and increasing the accessibility of health education material for older adults and family members and assisting in the documentation of care transitions.

The meeting was led by George Demiris, PhD, FACMI, Penn Nursing’s Associate Dean for Research and Innovation. “As artificial intelligence tools continue to advance, we have the opportunity to define what are the parameters for the use of AI in aging; rather than waiting to see how new tools may disrupt gerontology we have to examine how to facilitate appropriate, ethical, and transparent use of AI in aging. This meeting helps us think about the role we want AI to have in gerontology and identify challenges and implications as well as strategies to mitigate risks.” Demiris is also the Mary Alice Bennett University Professor, a Penn Integrates Knowledge Professor with joint faculty appointments in Penn Nursing and Penn Medicine, and he is also one of the principal investigators of PennAITech.

The discussion explored the following questions:

• Can Large Language Models be used to address social isolation and loneliness for older adults?
• What are guidelines to inform appropriate use of LLMs for systems that target older adults, persons with dementia, family members, clinicians and other stakeholders?
• How can we promote transparency in the design of LLMs for gerontology?
• How can system designers and evaluators address age-related bias (and digital ageism) in AI?
• How does ChatGPT affect the future of gerontological research?

The invited guests included experts in the fields of gerontology and geriatrics, LLMs, AI, and bioethics:

• Chris Callison-Burch, PhD, Associate Professor, Computer and Information Science, University of Pennsylvania
• Ryan Calo, JD, Lane Powell and D. Wayne Gittinger Professor, School of Law, University of Washington
• Jonathan H. Chen, MD, Assistant Professor of Medicine (Biomedical Informatics), Stanford University
• Jina Huh-Yoo, PhD, Associate Professor, Information Science, Drexel University
• Ari Friedman MD, PhD, Assistant Professor, Emergency Medicine, University of Pennsylvania
• Rebecca Krupenevich, PhD, AAAS Science & Technology Policy Fellow in NIH NIA’s Division of Behavioral and Social Research (BSR)
• Amanda Lazar, PhD, Assistant Professor, Information Studies, University of Maryland
• Marion Leary, PhD, Director of Innovation, School of Nursing, University of Pennsylvania
• Susan Reinhard, PhD, RN, FAAN, Senior Vice President and Director, AARP Public Policy Institute & Chief Strategist, Center to Champion Nursing in America and Family Caregiving Initiatives
• Steven E. Waldren, MD, MS, VP and Chief Medical Informatics Officer, American Academy of Family Physicians
• Anna Wexler, PhD, Assistant Professor of Medical Ethics and Health Policy, University of Pennsylvania

Other attendees included Penn Nursing PhD students Hannah Cho and Oonjee Oh, as well as Penn Nursing Staff Jennifer Forbes-Nicotera, Program Manager, and Nancy Greene, Data Project Manager. The group anticipates synthesizing the discussion to generate a white paper, by early Spring 2024, that will summarize key recommendations.