Document Type
Theses, Ph.D
Abstract
Conversational agents (CAs) have strong potential to support health and physical wellbeing through text-based coaching, but they remain limited in their ability to demonstrate supportive social behaviours that are important in human coaching interactions. In particular, relatively little is known about how Active Listening and Reassurance are perceived, modelled, and evaluated in text-based virtual healthcare coaching, or how such behaviours should be adapted to individual users.
This dissertation investigates how supportive interaction behaviours can enhance text based virtual health coaching, with an initial focus on Active Listening and Reassurance and a later theoretical emphasis on Active Listening. Across five unique studies, the research first examines whether users can perceive these behaviours in healthcare coaching dialogues, using a dataset of 135 dialogue excerpts drawn from real, handmade, and large language model (LLM)-generated sources. The findings show that users can perceive meaningful differences in supportive language across these dialogue types.
The dissertation then explores how Active Listening can be modelled computationally in a virtual coaching context using prompt-based techniques with LLMs. The results indicate that Active Listening can be generated as a controllable linguistic behaviour, with Retrieval Augmented Generation (RAG) producing the most effective responses among the prompting approaches examined. Building on this modelling work, a prototype virtual coach was developed and evaluated in a user study comparing an Active Listening version with a non-Active Listening version. The findings suggest that incorporating Active Listening can improve user experience, engagement, and outcomes related to behaviour change in text-based coaching. During the modelling and evaluation process, the research identified an important limitation in existing Active Listening models: most were developed for human–human and multimodal interaction and do not fully account for the constraints of text-based coaching.
To address this gap, the dissertation introduces LISTEN-R, a framework for conceptualising and operationalising Active Listening in text-based virtual coaching. Finally, the dissertation investigates whether users’ Big Five personality traits shape preferences for different levels of Active Listening. The results show that preferences for Active Listening intensity vary across users, indicating that supportive interaction should be personalised rather than treated as a one-size-fits-all design feature.
Overall, this dissertation contributes empirical evidence that Active Listening is both perceptible and consequential in text-based virtual coaching, provides a framework for operationalising it in this context, and demonstrates the importance of personality-aware personalisation in the design of more effective and socially responsive virtual healthcare coaches.
DOI
https://doi.org/10.21427/em2x-z598
Recommended Citation
Hussain, Ghulam, "Active Listening and Reassurance in Text-Based Virtual Health Coaches" (2026). Dissertations. 287.
https://arrow.tudublin.ie/scschcomdis/287
Funder
Research Ireland
Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License