The future of personalization in digital health

Background: As healthcare shifts from reactive treatment to proactive prevention, artificial intelligence presents unprecedented opportunities for personalizing digital health interventions across the lifespan. Recent advances in AI technologies promise to revolutionize longitudinal health management through continuous learning and adaptation. The ability to synthesize diverse data streams—from genomics to real-time behavioral and physiological metrics—enables increasingly sophisticated, personalized health models that evolve with individuals over decades. This focus session explores the potential of personalization for lifelong health optimization at the intersection of novel AI technologies and traditional human-centered design, examining both opportunities and challenges in this transformative approach to digital health.

Chair: Dr.-Ing. Jan Smeddinck, BSc, MSc

Speakers:

• DI Dominik Hofer, BSc
  Pre-Doc | Ludwig Boltzmann Institute for Digital Health and Prevention
• David Haag, MSc
  Pre-Doc | Ludwig Boltzmann Institute for Digital Health and Prevention

Abstract:

This talk will examine the potential of new AI systems for personalisation in digital health, with a focus on the capabilities of Large Language Models (LLMs) in the context of Just-in-Time Adaptive Interventions (JITAIs). We will begin by presenting the findings of our recent study, which investigated the viability of LLMs in this context. In this study, we used GPT-4 to generate Just-in-Time Adaptive Interventions (JITAIs) with the objective of promoting physical activity (PA) in outpatient cardiac rehabilitation. We conducted a comparative analysis of GPT-4 outputs with those from healthcare professionals and laypersons, evaluating the appropriateness, engagement, effectiveness, and professionalism of the produced JITAIs. The results demonstrated the significant benefits that can be achieved by utilising LLMs for the purpose of JITAI generation, while also offering further potential for effective personalisation. In the second half of the talk, we will provide an overview of the technical and implementation research aspects of using LLMs for personalised JITAIs in PA. We will present ongoing research efforts, including an open-source dataset focused on user acceptance of LLM-based JITAIs for model optimisation. Additionally, we will provide an overview of how context-aware computing can benefit from AI, leading towards context-aware AI systems.

Speaker Bios:

• Dominik Hofer studied AI and HCI, with a research focus on human-centered AI. He is pursuing a PhD at LMU Munich on “Investigating Context-Aware Artificial Intelligence for Dynamic Personalization in Digital Health” and works at the LBI-DHP.
• David Haag is a PhD candidate at the LBI-DHP, with a background in psychology. His research focuses on the momentary and motivational determinants of physical activity (PA), using ambulatory assessment methods. Building on the insights gained from his investigations, David designs and evaluates digital interventions, with a particular emphasis on Just-in-Time Adaptive Interventions (JITAIs), aimed at promoting increased engagement in PA.
  

Speaker:

• Dr. Rania Islambouli
  Post-Doc | Ludwig Boltzmann Institute for Digital Health and Prevention

Abstract:

As digital health shifts towards proactive and personalized care, recommendation systems are evolving into intelligent digital companions that support users throughout their health journeys. This talk explores the progression from traditional recommendation algorithms to adaptive systems that integrate real-time behavioral and physiological data, offering tailored support for a personalized and adaptive experience. We will examine how AI-driven, context-aware interventions can adapt to users’ changing needs, creating engaging and effective experiences. The talk will also address the challenges and opportunities in designing these digital companions to ensure they remain user-centered, accessible, and capable of promoting sustained well-being.

Speaker Bio:

• Rania Islambouli is a Post-Doctoral Researcher at the Ludwig Boltzmann Institute for Digital Health and Prevention. She completed her PhD in Computer Science at the Swiss Federal Institute of Technology in Lausanne, where her research focused on understanding user behavior and creating adaptive algorithms that deliver personalized recommendations, helping users make more mindful and satisfying choices online. Rania’s research interests include personalization and decision support in digital health technologies, as well as developing intelligent systems that adapt to users’ needs and preferences. She is particularly interested in how AI can be leveraged to create engaging and user-centered digital health experiences.

Speaker:

• Joseph Jay Williams
  Head of the Intelligent Adaptive Interventions Lab | Computer Science, Psychology & Statistics | University of Toronto

Abstract:

All 8 billion people have behaviours they want to change – to stop doing some actions, start doing others. Intelligent Interventions (delivered through apps, text, & people) can be conceptualized as finding the “Magic Words” to say that will help a person in a specific moment. This talk presents 4 techniques for personalizing and contextualizing these “Magic Word” Interventions using Adaptive A/B experimentation. We explain different kinds of Adaptive A/B experiments, and show how practitioners and scientists can use the AdaptEx framework (which received 1st in a $1M Xprize for novel experimentation techniques) to use Adaptive Experiments to perpetually enhance, personalize, and contextualize interventions. This involves integrating human intelligence (e.g. crowdsourcing from users, designers, & social-behavioural scientists) with artificial intelligence (e.g. LLMs & Reinforcement Learning), using both qualitative and quantitative methods. To learn more, see broader Research agenda (tiny.cc/williamsresearch), llist of >80 papers (tiny.cc/jjwcv), talk recordings (www.josephjaywilliams.com/representative-talks).

Speaker Bio:

• Joseph Jay Williams directs the Intelligent Adaptive Interventions lab at University of Toronto. He is an Assistant Professor in Computer Science, with courtesy appointments in Statistical Science, Psychology, Economics, Industrial Engineering, & the Vector Institute for Artificial Intelligence. The lab’s work is represented in over 80 papers (www.intadaptint.org/papers ), 2 Best Paper Awards (1 at CHI), 4 Runner Up/Honorable Mention for Best Paper (CHI, EDM, LAS), 1st place in a $1M Xprize competition for the future of experimentation technology in education, over $2M in grant funding, and interventions impacting over 500 000 people. Joseph is originally from Trinidad and Tobago, and has worked at the National University of Singapore, Harvard, Stanford, & UC Berkeley.

Speaker:

• Mike Trenell
  Prof. Hon. | Newcastle University & CEO | DAISER

Abstract:

The last 10 years has seen a rapid growth in digital health technologies. Yet, very few are used at scale. In this talk, Mike will explore what the future looks like and what is needed to get there, from the technical building blocks for making digital health, to the regulatory landscapes and how and who pays. A thoughtful reflection on a sector which needs to change how it is working in order to help people at scale.

Speaker Bio:

• Mike Trenell is a clinical scientist who retrained in technology. He has built 2 UK national digital platforms; the UK Health Innovation Observatory and UK Type 2 diabetes service. Mike is applying his 20+ year’s experience in technology into helping people build services in DAISER, a modular digital health platform.