AI Support for Health & Nutrition
Diet-related non-communicable diseases have become the leading cause of mortality globally, accounting for more deaths than non-diet-related mortality causes combined. We believe that machine learning and data science can support making health-beneficial decisions in our daily lifes. Artificial intelligance will contribute to faster and more accurate predictions of health-related situations, such as detecting arrhythmic heartbeats through your watch, sensing unhealthy purchase behaviour from your shopping data, or tailoring just-in-time adaptive interventions just right. Motivated through the advances of related technology and deep learning frameworks, we contribute to this field by piloting and assessing early-stage prototypes in the field of health and nutrition. To achieve that, we leverage existing data, novel hardware such as mixed reality headsets, and applied field studies under realistic real-world conditions.
- Exploring the potential of existing data assets such as ingredient information on products to support health behaviour interventions and monitoring of diets
- Just-in-time adaptive dietary interventions via embodied AI in mixed reality
- Using behavior tracking in mixed reality to trigger personalized interventions to nudge dietary behavior towards healthy behaviour
Current state of mHealth
Although thousands of diet and health apps exist, the vast majority of society is still not using them continuously. The design of mHealth apps has to improve dramatically & make use of late-breaking research to improve retention & effectivity.
With only little regulation by authorities on wellness- and diet-related mHealth (FDA.gov, 2019), users & dieticians suffer under a myriad of apps to choose from. Especially, diet apps suffer under self-selection of healthy users, short-lived retention & low adherence (Langford 2019), such that at least 50% of patients with diet-related diseases do not actively use them (König, 2017).
Barriers of usage among today’s diet apps lie in the extreme effort involved in manual diet logging, underreporting of diet, non-personalized & abstract recommendations, non-pleasant user experience. We recommend mHealth designers to include automated data collection methods, just-in-time-adaptive interventions, gamification, & state-of-art implementations of features, design, behavior-change-technologies.
Leveraging existing data
Due to the current diet apps fail to monitor, as more than 90% do not comply with logging their diets over more than 90 days. Furthermore, todays mHealth apps are perceived as non-personal, and often yield abstract, not directly actionable recommendations (e.g. eat less salt). For example, a lot of consumers do not know nutritional composition of often consumed products or do not know their daily average intake levels or exposure, nor do they know expected outcomes of their respective actions. Therefore, we suggest to use existing data such as purchase logs from loyalty cards or food composition data of relevant grocery products.
We can leverage such existing retail data, including image and ingredients, in order to improve contemporary mHealth applications on training health literacy, support continuous diet monitoring and triggering health related interventions. For example, users of a loyalty card can receive aggregate statistics on the nutritional composition of their recent groceries including helpful information on how to improve the dietary quality of the purchases if needed. Or wearers of mixed-reality headsets can be reached with effective just-in-time adaptive intervention right in the moment when relevant activities such as buying a snack, purchasing groceries or eating a meal are detected.
Find out more, by following our related projects in the field of AI Support for Health & Nutrition.