Human-like communication in conversational agents for delivering digital health interventions
Year: 2025
Type: dissertation
Abstract: As the global burden of non-communicable diseases (NCDs) continues to rise, healthcare systems are increasingly looking for adopting Digital Health Interventions (DHIs) to support behavior change and enhance chronic disease management. This thesis explores how human-like communication through Conversational Agents (CAs) can improve patient engagement, personalize care, and support sustainable self-management. Grounded in the Design Science Research Methodology (DSRM), this work integrates evidence from systematic and scoping reviews, user-centered design processes, and real- world pilot studies. It begins with a literature review identifying effective communication strategies and personalization mechanisms in DHIs. A scoping review follows, analyzing the automation techniques and human-like characteristics embedded in CAs across healthcare applications. These findings informed the development of a personalized, automated messaging platform aimed at supporting patients after cardiothoracic surgery. A solution was co-designed with clinicians, nurses, and patients and validated through iterative prototyping, a feasibility study, and a pilot trial at a public hospital in Lisbon. A pilot study was conducted with three patients for 1 month, which allowed assessing acceptability, system usability, perceived effectiveness, and improvements. The system achieved high levels of acceptability and usability and was considered a valuable tool to reinforce clinical recommendations during the post-discharge period by patients and the clinical team. Subsequently, a second iteration leveraged agentic LLMs to personalize behavioral interventions for NCDs in broader contexts, showcasing potential for generalizability. This solution was validated by physicians from a public health center. Key contributions include a framework for developing human-like CAs in healthcare, insights into personalization through behavioral and contextual factors, and the integration of AI for scalable communication. Limitations such as limited number of participants in the pilot studies and integration barriers with existing clinical systems are acknowledged, and future directions emphasize scalability, interoperability, and bidirectional communication. This work contributes to the design of secure, effective, and engaging digital interventions that enhance the continuity of care and empower patients in managing their health.
Author: Martins, Ana Catarina dos Santos
Language: English
Subfield: General Health Professions
Field: Health Professions
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7199338125