Organizing Chatbots for Health Symposium
We organized the Engaging Long-Term Interactions with Chatbots for Health symposium where we shared our progress and outcomes achieved over the past four years within the multidisciplinary Look Who’s Talking project. The symposium began with a short introduction by Tibor Bosse , and followed by three presentations, delivered by the project’s PhD candidates — Linwei He, Divyaa Balaji, and myself — who have been working collaboratively to develop and evaluate chatbots for smoking cessation and safe sex promotion. The event concluded with a panel discussion, driven by the audience’s thought-provoking questions, on the potential and challenges of implementing chatbots in healthcare. The symposium was held on December 10, 2024, in Nijmegen, The Netherlands.
Engaging Long-Term Interactions with Chatbots for Health
Trying to change a person’s health behaviour is difficult. The services offered by healthcare professionals is proven to be effective, but restricted. Chatbots, as a growing digital technology, on the other hand, hold great promise in supporting this process due to their accessibility, cost-efficiency, and scalability. By incorporating recent insights from the fields of social sciences, humanities, and computational linguistics, we have developed and tested a new generation of chatbots which can engage in long-term motivational interactions in two domains with different social dynamics and complexities; smoking cessation and safe sex promotion. The symposium aims to convey the findings and key insights acquired through the research and analysis conducted in the span of the last 4 years, contributing to the discussions in the fields of behaviour change, communication, human-computer interaction, and artificial intelligence.
Program
| 15 min | Opening remarks and introduction | Tibor Bosse | Radboud University |
| 45 min | Chatting Your Way to Quitting | Linwei He | Tilburg University |
| 45 min | Designing Chatbots for Behaviour Change | Divyaa Balaji | University of Amsterdam |
| 45 min | Leveraging Large Language Models | Erkan Basar | Radboud University |
| 30 min | Panel discussion | ||
Chatting Your Way to Quitting: Designing and Evaluating Chatbots for Long-term Smoking Cessation Support
by Linwei He , Tilburg University
Can a chatbot truly support long-term change, and is it feasible for people to stay engaged with this kind of digital tool? How do these conversations foster sustained user engagement and even a therapeutic relationship, and what is their impact on behaviour change, particularly quitting smoking? Through iterative design, we developed and evaluated multiple chatbot versions, focusing on feasibility, user experience, and intervention effectiveness. We examined how different conversational strategies influenced motivation and user experience over time. I’ll discuss key findings on these digital interactions and what they mean for the future of chatbots in health support.
Designing Chatbots for Behaviour Change: Evaluations of Two Novel Chatbot Concepts for Promoting Safe Sex Behaviours in Young People
by Divyaa Balaji, University of Amsterdam
Conversational agents are entering a new era of advanced computing and generative language models; we aspired to push the boundaries of how a conversational agent can support public health, and explored two new chatbot concepts might be used to promote safe sex behaviours in young people: one that targets individuals, and another that targets couples. Through qualitative methods, both chatbots were rigorously evaluated for user acceptance with the intent of understanding how and why users perceived the interactions in certain ways. In this talk, I will consolidate the findings from both evaluations, and discuss how they contribute to our understanding of user acceptance with conversational agents, as well as inform better design in future efforts.
Leveraging Large Language Models in Health-Promoting Chatbots in a Controlled Manner: A Human-Centered Approach and Insights
by Erkan Basar, Radboud University
Can task-oriented long-term chatbots be enhanced through the natural language generation capabilities of large language models like ChatGPT and LLama? How can their capabilities be harnessed in a controlled manner, particularly in sensitive fields such as health counselling? Our research explored these questions by developing a tool with an innovative dialogue manager tailored to integrate these models’ generated utterances into predefined dialogue structures, aiming for controlled and effective communication. By conducting user studies with real conversational data from chatbots focused on smoking cessation and sexual health, we obtained insights into how users perceive the responses generated by large language models. I’ll share key findings on how these digital dialogues shape user engagement and reliability in health counselling scenarios.
Images
Find and like this news on LinkedIn