A2004 Personalised Recommender |                  Technology Maturation | KULeuven

PERSFO is designing and testing an innovative, artificial intelligence-driven, personalised recommender platform, with the primary aim of influencing healthy eating patterns in the workplace.

The platform will generate personalised dietary advice for consumers using both catering facilities and convenience food delivery services (vending machines) in workplace environments. Individuals will receive personalised dietary advice via smartphone technologies, based on food service providers’ menus in their workplace using personalised behavioural motivators, i.e. some people want and need direct instructions whilst others prefer hints and nudges that remind them gently to eat healthily.

Consumers will receive real-time information via PERSFO on food options provided by their food service that are most suited to their health goals or dietary preferences/needs, whilst still allowing them chose what to eat. This approach is unique and beyond the current state-of-the-art.

More specifically, the PERSFO recommender engine will deliver personalised meal suggestions, designed based on three aspects, namely overall dietary intake, health parameters, and persuasion profiles using different levels of information density. For diet, information might be highly detailed through manual food logging or more general based on information logged automatically for at-work purchases. Health information could be simple biometrics or more detailed data using diagnostic tools and anthropomorphic data. And, finally, persuadability will be determined using either full psychological survey assessments or based on only logged interactions with the PERFSO platform.

PERSFO is also exploring behavioural change in canteens, combining the recommender engine with personalised dietary advice to whether personalised advice and motivational recommendations can affect engagement and behaviour around food choice. Based on data from diet, heath, and persuasion profiles, as well as declared goals and needs of users, different behavioural change approaches will be deployed, such as: nudging, manipulating choices, group motivation, visualisation, and gamification, to increase consumer acceptance, engagement and compliance with healthy eating goals.

Currently, access to provider APIs (application programming interface – a set of functions and procedures allowing apps to interact with operating system and other service) is complete but Covid-19 public health controls mean data acquisition and pilot testing will take place simultaneously in late 2020. Similarly, whilst the first design concepts for the recommender are available, the full version will be implemented in late summer/ early autumn and completed at the end of the year.