A chatbot that enables self-reflection on sleep disturbance factors.
Personal Final Master Project.
Mentor: Yaliang Chuang. Oct. 2019 – Jun. 2020. Future Everyday, TU/e
As sleep health is regarded as important in recent years, many wearable products and mobile apps have been developed for users to track their sleep and interpret sleep quality. Most of the available solutions have mainly focused on objective measurements, such as body movement or bedroom temperature. However, due to the lack of users’ subjective measurements, sleep trackers failed many users in identifying their sleep factors and making connections to modifiable behaviors and sleep hygiene. In this study, we developed SlumberBot with conversational chatbot technology to help users capture subjective experiences of sleep and the relevant factors in daytime activities. SlumberBot uses a semi-automated approach by connecting to the objective data collected with a Fitbit wristband and utilizes an interactive chatbot to gather users’ subjective experiences in relation to the Fitbit data. Through a 5-minute conversation in the morning, SlumberBot can proactively initiate the inquiry of a user’s behaviors on sleep-related factors and generate corresponding reports.
SlumberBot was developed on Google’s Dialogflow and integrated into Facebook Messenger. It was designed to be used by users upon their awakening every morning. The interactions between users and SlumberBot consisted of two sessions: i) conversation session and ii) sleep report session.
Screenshots of two sessions
In the conversation session, the chatbot provided different questions based on the user’s actual sleep condition, which is collected from their Fitbit account via OAuth protocol. We designed the dialogue based on the Consensus Sleep Diary (CSD), as it contains a defined set of questions to collect objective and subjective items to log for sleep analysis. The conversation started from querying users’ subjective sleep quality (i.e. “How was your sleep last night?”). Based on users’ responses to the first question and their detected sleep condition, SlumberBot selected most relevant topics to proceed the conversation (e.g., For the users who had a poor sleep, it asks more about their sleep problems such as awakenings, sleeplessness, early wake-up, and some contributing factors, whereas it asked other questions for the users who had a good night’s sleep.) Under each specific topic, a series of questions was carried out for inquiring about contextual factors to prompt reflection. These questions focused on asking two main underlying reasons for each sleep-related problem and the degree of impacts of those factors on their sleep quality. To reduce users’ input burden, some of the common factors (e.g. noise, stress and anxiety, temperature, light, etc.) were provided to users in quick reply options.
Conversation flow of the SlumberBot
Example of chatbot questions under on specific topic
Example of quick reply options
At the end of the conversation session, SlumberBot generated a daily sleep report to summarize users’ last night’s sleep. In the sleep report session, SlumberBot analyzed and compared the impact of recorded factors. SlumberBot visualized users’ sleep factors in the last three days and provided several tips to guide users’ sleep hygiene. We used tag-clouds to help users visually see the factors that positively or negatively affected their sleep. The sizes of the tags are set based on users’ replies on the factor impact questions.
Example of the daily sleep report
Field User Study
With SlumberBot, we conducted a 4-week user study with five participants. The goal of this study was to understand overall self-tracking experience with SlumberBot and to examine how SlumberBot could support users’ self-reflection on their sleep. The result shows that SlumberBot is easy to stay engaged with and supportive in users’ reflections on personal contextual factors. Besides, SlumberBot has shown the potential of triggering short-term behavior changes that are easy to achieve in daily life. Further, from the study, we also learned some future directions for improving the SlumberBot design and conducting the follow-up works.
Some highlighted quotes
Because the background of this study, sleep health, was a new direction that I hadn’t explored before, I actually experienced many challenges throughout the whole study. At the same time, these challenges also benefit me in many aspects.
Firstly, this project helped me develop a good habit of being practical. For example, in the development process of SlumberBot, I kept the habit of making quick decisions and focusing on outputting practical works. I have become more productive in design than before.
Secondly, although this project was launched under a brand new topic, I still got the chance to use the skills that I learned from previous courses and projects. The project process trained and strengthened my ability on these skills, and let me find my deficiencies. For instance, I’ve found that my ability to organize semi-structured interviews needs to be improved, so that I could have a ready-mind to come up with extra questions that were outside the outline but relevant to users’ responses.
Thirdly, since this project involves a lot of directions that I have not tried before, it actually helped me enrich my design skills in my expertise areas. For example, now I had the experience of a project that focuses on the user experience of digital design, which is not included in my previous works. I believe that these improved abilities can help me earn some good jobs in my future career.
As for the missed opportunities, the project has not kept up with the expected schedule. I was planning to finish the field study in April so that I still got one extra month to complete a final iteration of the prototype. Actually, the Google Dialogflow platform I used is also able to get connected with Google Home and Amazon Echo, so it was planned that maybe we could try to apply this prototype to other modalities to explore its effect. However, now we have to put this plan in future works.