Precision circadian and sleep health: Modifying biologically-relevant light exposure through smart conversational agents
Project Summary:
Exposure to natural light triggers a range of neuroendocrine, behavioural, and cognitive responses, including the synchronisation of the circadian clock with the light-dark cycle and modulation of alertness. Due to the fundamental link between light exposure and physiology, light can both improve and disrupt our health and well-being. However, many people are unaware of their light exposure and its consequences on wellbeing, circadian health, and sleep, and how their behavioural choices modify light exposure. Furthermore, current interventions to improve circadian health and sleep are not tailored to individual needs. To address these current limitations and move towards a “precision health” approach, low-access strategies to educate the public are warranted which also provide simple prevention tools and offer first help for those with early signs of circadian health-related difficulties.
Digital solutions, such as chatbots, are a promising way forward to achieving these goals. Chatbots are applications on messaging platforms (e.g., Telegram or WhatsApp) that mimic text-based human interaction by automatically providing content and personalised responses user. Here, we will examine the feasibility of chatbots to promote personalised circadian and sleep health. This includes sampling the user’s geolocation and local weather data to provide individualised recommendations on healthy light exposure for the upcoming day(s) with a Telegram based chatbot system which was first developed by the 2023 TUM PREP students. This chatbot system is also used to sample baseline user characteristics upon which individual recommendations are based. Baseline user characteristics include age, gender, and chronotype but will be extended to household and civil status characteristics (e.g., children, pets, spouses) as well as lifestyle preferences and social schedule constraints (e.g., sports or timetabling issues).
We will further develop the existing chatbot to run a micro-randomised controlled trial to support just-in-time adaptive interventions (JITAIs). JITAIs are used in digital health to provide timely and personalised support to individuals based on their current context, needs, and behaviour. These interventions leverage real-time data and technology to deliver appropriate assistance when it is most relevant and effective.
Prior to the student’s arrival, the project is planned to be in the following stage as prepared by the study team and the 2023 TUM PREP student cohort:
- Telegram chatbot infrastructure built
- Ethics application for a micro-RCT submitted and approval granted
Individual student tasks:
Student 1 (informatics/computer science background):
- Implementation of a framework to implement micro-RCTs in an existing chatbot platform
- Incorporating missing JITAI components in the system including e.g. real-time monitoring, contextual awareness (inclusion of context factors such as time of day or social settings), personalisation, adaptability, feedback and behavioural change techniques
- Incorporate AI models to improve communication with the user
Student 2 (psychology/(bio)medical/behavioural/cognitive science background):
- Trialling the beta version of the chatbot with human participants
- Development of intervention modules incorporating evidence-based techniques such as SMART goal setting and monitoring via diary input, habit creation and/or behavioural change techniques
- Setting up and running a micro-RCT as a feasibility study
- Collecting data and developing a first statistical analysis of the data
General Tasks for both students:
- Documenting upcoming technical and other errors
- Overviewing data collection and first data analyses
- Collecting meta-data and user behaviour
- Writing trial reports identifying current challenges and next steps
- Suggesting improvements to the chatbot
- Summarising learnings in a short report or presentation
- Participating in team meetings
Expected outcomes of the project:
- Functional chatbot implementation incorporating JITAI/micro-RCT components to form the basis of a larger data collection effort investigating longitudinal light exposure across seasons
- Report on the challenges of implementation of a JITAI/micro-RCT chatbot infrastructure, to be co-authored with the supervisors
- Data set collected with the JITAI/micro-RCT chatbot, forming the basis of an article describing the feasibility of this approach
