1. General
Category
SDG 3: Good Health and Well-being
Category
Computer Software
2. Project Details
Company or Institution
Ghent University
Project
A Motivational and Personalized Recommender Systems for Healthy Physical Activities
General description of the AI solution
The project is about a motivational and personalized Recommender System for healthy physical activities. The research combines recommendation algorithms from computer sciences with evidence-based Behavior Change Techniques from motivational psychology. The motivational goal is to make sure the user is motivated to use the system and has the motivation to keep executing the recommendations for physical activities. The personalization goal is to get to know the preferences for physical activities of the user as fast as possible and to make sure the personalized recommendations are accurate for that user, but also to discover new preferences for the user and to go beyond current preferences.
Innovatively integrating mood and context into the system, the Recommender System knows which activities make the user feel good and in which context (such as the weather, the location and the location type, e.g. in nature, in a building or on a road), and it focusses on generating recommendations for activities that are good for both the physical and mental health of the user. The system will be tailored to the user’s current health level, gradually increasing the recommendations towards more healthy behavior over time. All of this has the aim to increase the motivation of the user to have more physical activity, accomplish sustainable behavior change towards more healthy behavior, establish long-term adoption of the suggested activities and eventually improve the overall health of its users, both physical and mental health.
Website
Organisation
WAVES, Ghent University
3. Aspects
Excellence and Scientific Quality: Please detail the improvements made by the nominee or the nominees’ team or yourself if your applying for the award, and why they have been a success.
The AI solution is based on the Machine Learning techniques of Recommender Systems (RSs). RSs are nowadays widely used in streaming services, online stores and social media to predict the users’ preferences and make personalized suggestions. The RS in this research will create personalized suggestions for physical activities (e.g. walking for 30 minutes, running for 20 minutes) and will go further than predicting behavior by also surprising the user with new, healthier activities and breaking old, unhealthy habits.
The RS will combine algorithms from both content-based (recommending items similar to items preferred before) and collaborative filtering (uses other, similar users’ preferences). The RS starts with small, easy activities, gradually improving towards more healthy, more difficult behavior. To expand the user’s existing preferences, the RS will implement an algorithm based on random walks with restarts in a graph of activities. By using a weighted graph where the weights of the links between them are oriented towards more healthy and difficult activities, users will be guided towards them, gradually transitioning, starting at their personal level and considering their personal preferences and context.
Active research and development of the RS are initiated (TRL 3, TRL 4 when the current observational study with the first app is finished). Extensive literature review has been conducted in both fields of RSs and motivational psychology. Elements of the technology are being studied by collecting observational data through a first mobile app: people’s context, mood and physical activities in their day-to-day life. Collecting these data shows the feasibility to collect it with an app, their influence on each other and provides the content for the recommendations (e.g. which activities can be recommended in which context and mood). The results of this research will be presented at conferences and published in A1 journals in the domain of computer science.
Scaling of impact to SDGs: Please detail how many citizens/communities and/or researchers/businesses this has had or can have a positive impact on, including particular groups where applicable and to what extent.
There is no doubt that improvements can be made in people's health in today’s society. Noncommunicable diseases are the world’s leading cause of death, reduce people’s life quality and cause big costs on health care. The World Health Organization wants to eliminate their avoidable burden and emphasizes their behavioral and metabolic risk factors such as physical inactivity, unhealthy diet, obesity and raised blood pressure. In Belgium, 28.8% is at the risk of compromised health due to a lack of physical activity and almost half of the people is overweight (BMI ≥ 25). Studies regarding the coronavirus conclude that obese patients are more likely to progress to severe COVID-19 and are recommended to practice mild-to-moderate physical exercise to improve the immune system. Healthy behaviors in physical activity contribute to both physical and mental health and may reduce negative health outcomes.
The RS proposed here will open doors to more effective and personalized recommendations for healthy behavior because it takes into account motivational aspects of physical activity, people’s preferences for physical activity and the influence on people’s mood, benefiting both physical and mental health, which leads to healthier lives and better wellbeing of the population for all ages, as described in SDG 3. This way, people will be more motivated to perform the healthy behavior long-term, decreasing preventable behavior related illnesses and financial burdens on the economy and on businesses.
The project would eventually result in new methods for RSs concerning the health of people and their motivation to perform healthy activities long-term, taking into account their preferences, context and mood. The project would combine the motivational aspect of psychology with the personalization aspect of RSs, both working together, closing the gap for researchers of existing systems and contributing to the long-term adoption of a healthy lifestyle.
Scaling of AI solution: Please detail what proof of concept or implementations can you show now in terms of its efficacy and how the solution can be scaled to provide a global impact ad how realistic that scaling is.
The RS will be implemented in a mobile application to deliver the recommendations anywhere the user goes. Mobile apps can be distributed worldwide and the user community can grow to everyone who has access to a smartphone and the internet. The network effects are also clear: the more adopters of this technology there are, the more user information there is available for the collaborative filtering algorithm and the more information about different activities there is available (as content for the recommendations, connected to more contexts and mood values), both leading to an increase of the utility value for each user: the RS can discover new user interests, generate more accurate recommendations, motivate the users more to keep engaging in the physical activities and establish sustainable behavior change. But even for the first users (cold start problem), the system will be capable to generate meaningful recommendations because of the content collected through the observational study with the first app, linked to context and mood, as described above.
Motivating people to engage in more physical activity does not only improve the physical health of that person, but also reduces that person’s ecological footprint and amount of carbon dioxide emissions from transport (e.g. by replacing a car ride to the local grocery store with a bike ride). The RS will learn the preferences and lifestyle of its users in order to create sustainable behavior change and this includes finding ways to integrate the physical activity into day-to-day tasks of that person.
RSs and Machine Learning are hot topics nowadays and implementing the technology with our own collected dataset of physical activities and our algorithms, as described before in this submission form, will be feasible with the expertise of the research group WAVES of the Department of Information Technology of Ghent University.
Ethical aspect: Please detail the way the solution addresses any of the main ethical aspects, including trustworthiness, bias, gender issues, etc.
The mobile device can be used as a medium for the healthy activity recommendations and it has the potential to reach a lot of people worldwide. The smartphone app can be carried everywhere the user goes, deliver timely notifications and interact with its user.
Because the system works with sensitive personal data like location data, physical and mental health data, the feasibility of the data processing in the RS has already been extensively studied in terms of the GDPR. With the help of Data Protection Officers and lawyers, legal advice (e.g. encryption of the datasets and pseudonymization) and legal documents (e.g. Terms of Use and Privacy Policy) have been provided for this research in order to be GDPR compliant through all the stages of the research. Ethical approval for the study has been received from the ethical committee of Ghent University. When working with Machine Learning and AI, careful attention will be made to prevent bias from creeping into algorithms by training the system anonymously (e.g. without gender as a variable). Because the system has access to subjective mental and physical health data of its users, it is of good practice to alarm its users timely to visit a doctor or professional when necessary, to remind its users that the AI is not a substitute for medical or professional care and to keep discussing and implementing necessary and possible security measures with professionals.