Review mining to discover user experience issues in mental health and wellbeing chatbots

Abstract

Abstract: Mental health and wellbeing chatbots are growing in popularity. Involving the end-user in creating these products is an important design consideration, to ensure technologies meet user needs and are easy to use. Extensive databases of app reviews provide rich data sources which can inform design, based on user feedback of apps already in existence. In this study, review mining was conducted on app reviews (n=20,461) across 7 mental health and wellbeing chatbots, focusing on the reviews that included the topics of design and user experience. The aim is to establish what user experience issues of mental wellbeing chatbots can be discovered. Natural language processing techniques were used to analyse reviews, and k-means clustering was applied to identify similar reviews based on content. These processes can be used to provide recommendations to designers of digital mental health technologies.