This dissertation presents novel theoretical and technical frameworks and offers fundamental insights to advance chatbots towards facilitating responsible, healthy, and engaging conversations for promoting a healthy lifestyle. The integration of chatbots in health behaviour change interventions presents both opportunities and challenges. Chatbots offer a promising and scalable way to deliver personalized, immediate, and accessible support. Yet, conventional approaches are often limited by restricted and repetitive conversations, leading to low user engagement and users abandoning interventions prematurely. Recent generative language models bring greater flexibility and variety to dialogue, allowing for more adaptable and personalized interventions, but their unpredictability risks causing miscommunication. Ensuring that chatbots are implemented ethically, confidentially, and without bias is imperative when these chatbots are designed to support health behaviour change. Therefore, the goal of this dissertation is to develop and evaluate engaging chatbots with adequate conversational abilities for promoting a healthy lifestyle. It outlines four main contributions: a theoretical framework for an ethical implementation that respects user autonomy; a chatbot development framework that combines generative language models with structured strategies to sustain engagement while preserving control; empirical evidence that generative language models enrich interactions yet require human oversight; and findings that chatbots can provide consistent support to stimulate healthier behaviours.
Erkan Basar obtained his Bachelor’s degree in Computer Engineering from Muğla University in Turkey in 2015 and his Master’s degree in Artificial Intelligence from Radboud University in 2017. From 2017 to 2020, he worked as an AI engineer at FloodTags, where he developed information-processing systems to capture critical information about natural disaster events in real-time from online media. In 2020, Erkan started his position as a PhD candidate at Radboud University, which resulted in his dissertation on developing and evaluating engaging chatbots that support health behaviour change. Since 2022, alongside his doctoral research, he has been actively involved in the Kon ta bAI project, which focuses on developing AI models for the Papiamentu language. He is currently appointed as a postdoctoral researcher at Wageningen University.