Hotline Bling: Chatbot M.A.X. that provides hotlines based on the user's mental health

Thumbnail Image
Issue Date
Journal Title
Journal ISSN
Volume Title
Anybody can go through a hard time. It has been estimated that 29% of the people worldwide su er from mental health disorders at least once in their lifetime (Abd-Alrazaq et al., 2019). By sharing your feelings, it could lighten the negative state of mind a bit. Nevertheless, not everyone is able to talk to a therapist. This includes several reasons, such as shame or lack of money and/or time. Although mental health chatbots cannot function as a replacement (yet), they can be useful as an additional tool. Because con- versations about mental health can often be quite heavy, it is important a chatbot is able to recognize the user's mental states (detect) and provide the appropriate hotlines1 if needed (react). This thesis forms a pipeline2 with the thesis of M. Reksoprodjo. Her thesis focused on the detection part, while mine focused on the reaction. However, note that the actual implementa- tion of the pipeline could be done in further research. The research question concerned whether it was possible to build a working non-embodied chatbot that detects SOS signals of poor mental health and is able to refer the user to hotlines. Literature research was done and a proof-of-concept chatbot, called M.A.X. (Mental AI Access), was created to answer the research ques- tion. In addition, a web-application was made to deploy the chatbot on. M.A.X. was reviewed along the following scienti c measures: e ciency, ef- fectiveness, dialogue handling, anthropomorphism and esthetics. While the chatbot gives hotlines when triggered to do so, further research is needed to draw strong conclusions and answer the research question properly.
Faculteit der Sociale Wetenschappen