Coproducing multilingual conversational scripts for a mental wellbeing chatbot - where healthcare domain experts become chatbot designers

Introduction: Digital mental health interventions, such as chatbots that promote mental health and wellbeing are a promising way to deliver low-threshold support 24/7 for those in need. According to current knowledge about the topic, health care professionals should participate in the design and development processes for digital interventions.
Objectives: The aim of this presentation is to describe the interdisciplinary content development process of the ChatPal chatbot.
Methods: The content development process started in co-operation with mental health professionals and potential users to identify requirements, and to outline how the chatbot’s ‘personality’ should be designed. A theoretical model was used as a framework for activities and conversations with the chatbot. Conversation scripts were created, evaluated and tested in international, multi-disciplinary group workshops including mental health professionals, interaction design researchers, healthcare scientists and computer scientists. Initial conversational scripts were drafted in English. All content was then translated into other languages (Finnish, Swedish and
Scottish Gaelic) and cultural adaptation took place during the translation process. Online tools were used to allow international users to collaborate and co-produce the chatbot scripts.
Results: A multilingual chatbot was developed and the conversation scripts were structured and stored using a spreadsheet. The conversation scripts will be made freely available online in due course using this structured approach to formatting chatbot dialogue content. Making this openly available will allow organisations and researchers to repurpose the content as well as facilitating studies that wish to assess the design of conversation scripts for mental health chatbots. The conversation scripts were reviewed and edited during the design process however, these scripts still required modification so they could be implemented in the chatbot. This highlights the challenges in
turning empathetic and supporting conversations to short utterances suitable for a chatbot.
Conclusions: The ChatPal chatbot is now available in four languages. As literature about the topic is still scarce, it is important to describe and document the content development processes of mental health chatbots to outline best practices and to foster mental health professionals' role and increase their knowledge about chatbot development processes. Future work will develop a conversational UX toolkit that would allow health professionals to design chatbot scripts using design guidelines and principles.