Abstract
We consider the development of reinforcement learning (RL) algorithms for mobile health interventions in which the unit is a dyad and the intervention has multiple components each involving sequential decision making. Different components target different members of the dyad and/or their relationship. Sequential decision making for the different components is often at different time scales and further each component may be designed to impact different causal chains leading to the primary health outcome. To optimize the effectiveness of a multi-component digital intervention, we have developed a Multi-Agent Reinforcement Learning (MARL) approach. By incorporating domain knowledge, the MARL approach in which each agent is responsible for the delivery of one intervention component, is able to learn faster compared with a flattened agent. This work is motivated by our involvement in an upcoming trial using RL to personalize a multi-component mobile health intervention, ADAPTS-HCT. ADAPTS-HCT is designed to improve medication adherence by individuals who have undergone hematopoietic cell transplantation.
Presenter
Susan A. Murphy is Mallinckrodt Professor of Statistics and of Computer Science and Associate Faculty at the Kempner Institute, Harvard University. Her research focuses on improving sequential decision making via the development of online, real-time reinforcement learning algorithms. Her lab is involved in multiple deployments of these algorithms in digital health. She is 2013 MacArthur Fellow, a member of the US National Academy of Sciences and of the US National Academy of Medicine. She is a Fellow of the College on Problems in Drug Dependence, Past-President of Institute of Mathematical Statistics, Past-President of the Bernoulli Society and a former editor of the Annals of Statistics.
