Overview: Generative Artificial Intelligence (or Gen AI) refers to a class of artificial intelligence systems with the ability to produce text, images, video and/or audio based on prompts provided by a user. Well-known Gen AI services available to the public include Open AI’s ChatGPT, Google’s Gemini and Microsoft’s Copilot. Gen AI systems can also be effective at analyzing and summarizing text and image data. Gen AI systems learn these abilities by processing large amounts of data through computational models using methods from the field of machine learning. Research on Gen AI models is ongoing, including extending their abilities, reducing the occurrence of incorrect responses (often called hallucinations), and improving other aspects of safe use. Current Gen AI models are being actively deployed in a diverse array of areas within health care, including in support of aging and AD/ADRD research.
MassAITC Pilot Project Highlights: MassAITC has funded multiple projects making careful and safe use of current Generative AI models in different problem areas including improving the accessibility of clinical trials, helping to protect older adults from phishing attacks, and improving support for caregivers. In 2024, MassAITC pilot grant awardee and Boston area startup Kinto was acquired by Rippl Care. MassAITC Year 3 pilot awardee Dr. Gang Wang (UIUC) published and presented results from their pilot study on detecting text message scams at a renowned conference, the Twenty-First Symposium on Usable Privacy and Security (SOUP). More information on funded pilots in this area is listed below, along with additional resources including MassAITC webinars touching on this topic area.
Empowering Caregivers of Individuals with Cognitive Impairment to Make Safe Nonprescription Drug Decisions
Eun Kyoung Choe, University of Maryland-College Park. This pilot develops Aidara, an AI-powered digital health system that helps caregivers of individuals with cognitive impairments make safer over-the-counter medication decisions. Through multimodal interaction and personalized guidance, Aidara aims to support informed decision-making and reduce medication-related risks.
Scalable LLM Facilitation for Peer-to-Peer Support Groups of Informal Caregivers
Gregory Stock, Socratic Sciences Inc. Hamed Zamani, University of Massachusetts. Mary Mittleman, New York University. Socratic Sciences, in collaboration with UMass and NYU, is developing an AI-facilitated, question-driven app to help family caregivers of those with AD/ADRD connect and support one another in small, trusted groups around open-ended questions. The project will test and refine a group-facilitation AI bot to make meaningful peer support scalable, accessible, and affordable for millions of overwhelmed caregivers.
Neuroimmune Foundation Models for Uncovering Biomarkers in Alzheimer’s Disease and Related Dementias
Mariano I. Gabitto, Allen Institute. This project seeks to develop and train a novel machine learning foundational model that unifies brain and peripheral immune system omics data to identify blood biomarkers and map cellular changes in AD/ADRD.
Leveraging AI for Just-in-Time Smartphone Solutions for Family Caregivers
Felipe A. Jain, Massachusetts General Hospital, Finale Doshi-Velez, Harvard University. Family caregivers of people living with dementia have high needs for skills training and methods to reduce stress. This project will study the feasibility of a just-in-time adaptive intervention delivered by smartphone to increase engagement and helpfulness of caregiver skills and relaxation content for caregivers.
TRIALCHAT: Leveraging LLMs to enhance AD/ADRD clinical trial participation
Tim K. Mackey, S-3 Research LLC, Joshua Yang, California State University, Fullerton. This project will aim to develop TrialChat, an AI-powered chatbot and clinical trial navigator designed to increase participation in Alzheimer’s disease and related dementias (ADRD) clinical trials by providing tailored education, personalized trial matching, and recruitment support for older adults and caregivers.
Using AI to Repurpose Small Molecules to Target Amyloid-tau Interactions in AD
Jeremy Linsley, Operant BioPharma. This project will use a proprietary robotics and artificial intelligence to identify targets and clinically-tested drugs that could be repurposed for Alzheimer’s disease by blocking the harmful interaction between Tau and Amyloid Precursor proteins.
Protecting Patients against Phishing Attacks using AI-enabled Agents
Gang Wang, University of Illinois at Urbana-Champaign. Roopa Foulger, OSF. This project will design, prototype, evaluate, and potentially deploy an AI-enabled voice agent to assist patients (especially older adults) to better recognize phishing messages and reduce cybersecurity risks during patient outreach and communications.
An AI-powered Digital Therapy Assistant for Monitoring and Treating Cognitive Impairment
Jennifer Flexman, Moneta Health. Michael Busa, UMass Amherst. This project will develop AI algorithms used by Moneta™ digital therapy assistant to monitor the speech of individuals with mild cognitive impairment and early dementia during cognitive rehabilitation therapy.
Creating a Framework for Large Language Models for Caregiver Support in Dementia
Ipsit Vahia, Rachel Sava, McLean Hospital. Joseph Chung, Rippl. The CAREALL pilot project explored the use of large language models (LLMs) to support caregivers of individuals with dementia.
Intelligent Cognitive Assistant for the Individuals with AD/ADRD for Handling Word-finding Difficulty
Archna Bhatia, Institute for Human & Machine Cognition, George Sperling, UCI. This project aims to develop an Intelligent Cognitive Assistant that provides real-time word retrieval support as well as personalized training to enhance word retrieval for individuals with mild cognitive impairment and Alzheimer’s Disease and Related Dementias.
Application of Sentiment Analysis and Generative Language Algorithms to Kinto, a Support Service for Family Caregivers of Persons Living with ADRD
Joseph Chung, Kinto. The Kinto pilot project explored how AI technologies—specifically sentiment analysis and generative language models—could enhance support for family caregivers of individuals living with Alzheimer’s and related dementias (ADRD).
MassAITC Webinars on Generative AI
Upcoming Webinar – Old School Meets New School: Voice-Based, AI-Enabled Cognitive Rehabilitation for Dementia Care
Zoom Registration: https://umass-amherst.zoom.us/meeting/register/o-oU8bqwR4GnY7D0cDLaeg Abstract: Cognitive rehabilitation therapy supports individuals living with mild cognitive impairment and early-stage dementia in maintaining and improving function in daily life. However, access remains limited due to constraints in the availability and scalability of trained therapists. Recent advances in artificial intelligence, combined with evolving reimbursement pathways for remote care in the United States, now make virtual delivery models increasingly viable, creating new opportunities to expand access to high-quality cognitive care. Moneta Health has developed a telephone-based cognitive rehabilitation platform that enables structured, personalized therapy sessions delivered remotely and overseen by licensed speech-language pathologists. The platform leverages AI-driven speech analysis and automated session orchestration to support consistent therapy delivery while preserving clinician oversight, enabling older adults to engage in care from their homes through a familiar and
Past Webinar – Can You Walk Me Through It? Explainable SMS Phishing Detection using LLM-based Agents
Abstract: Phishing attacks pose a significant threat to users, especially older adults. Existing defenses mainly focus on phishing detection but often cannot explain to lay users why a message is malicious. In this talk, I will discuss how we use Large Language Models (LLMs) to detect SMS phishing while generating evidence-based explanations. The key challenge is that SMS is short, lacking the necessary context for security reasoning. We develop a prototype called SmishX that gathers external contexts to augment the chain-of-thought (CoT) reasoning of LLMs and facilitate the explanation process. I will further discuss our user studies to evaluate the effectiveness and usability of SmishX. Finally, I will discuss the open challenges and opportunities of using AI to help older adults better protect themselves from cybersecurity threats in general. Biography:
Past Webinar – Progress in Personalizing Content and Dosing of a Physical Activity Promotion Intervention, David E. Conroy
Abstract: The Michigan Roybal Center aims to develop physical activity interventions for middle-age and older adults that engage validated mechanisms for adhering to behavior change following the end of active intervention support. This talk will review our ongoing work (a) to develop person-specific dosing algorithms to select the content and timing of text messages and (b) to engineer prompts for generative artificial intelligence systems to author message content that activates affective motivational processes to promote physical activity. The long-term objective of fusing these personalization strategies is to improve adherence to behavior change and reduce risk for Alzheimer’s disease and related dementias. Biography:
Past Webinar – LLMs for Assistance in ADRD: From Word Retrieval to Caregiver Support, Archna Bhatia and Richard Curtis
Overview: This webinar comprises two presentations by Archie Bhatia from the Institute for Human & Machine Cognition and from Richard Curtis from Ripple Care. They each discuss the work from their MassAITC a2 Pilot awards using LLMs to help with word retrieval for older adults with ADRD and to support ADRD Caregivers. Abstracts: About the Speakers: