Conversations with VA
Preventive Mental Healthcare
through Conversational User Interfaces
Master Thesis Project| 10 months
Department of Industrial Engineering & Innovation Sciences, University of Technology Eindhoven, NL
Supervisors: Prof.Dr. Wijnand Ijsslesteijn, Dr. Minha Lee
​
Skills: Experiment Design, Data Analysis (STATA), Python, Angular, Conversation Design, Dialogflow, HTML, CSS
Abstract
Voice User Interfaces have become very prevalent in our lives. While they make our lives easier by assisting with daily tasks, they can also be developed to help cater to our emotional needs. In distressing times like the COVID-19 pandemic, they can provide preventive and interventional therapy and improve mental resilience. They can do so by stimulating self-compassion, a concept that can alleviate feelings of loneliness and also combat mental health illnesses like depression and anxiety. The current study investigated the effect of a gender-ambiguous voice user interface on self-compassion by providing or asking for care. The participants (n=161) engaged either in conversation with a voice
user interface called ’VA’ that provided or asked for care, or completed self-compassion improving tasks in online questionnaires. Overall, neither improvement in self-compassion nor reduction in perceived loneliness were observed. The care-giving VA was perceived to be more female-gendered. Interestingly, this is in line with stereotypical gender bias regarding male-female differences: females are perceived to be (and expected to be) more caring and nurturing than males. The limitations of the current study discuss potential pitfalls and suggestions to consider for future research involving voice user interfaces.
​
Keywords: Mental healthcare, self-compassion, well-being, positive psychology, voice user interfaces, conversational user interfaces
The motivation for this research project was presented as a position paper at the Conversational User Interfaces (CUI) Workshop at the Computer Supported Cooperative Work (CSCW) 2020 conference.