A personalised, goal-driven, and explainable chatbot for nutritional coaching based on a multi-agent system

Abstract

Obesity and related sicknesses (e.g. type 2 diabetes) are global, ever-growing public health problems (Chooi, Ding and Magkos, 2019). In 2020, in particular, obesity had gained further significance since people who suffer from it have a higher likelihood of a severe disease progression after contracting SARS-CoV-2 (Zhou, Chi, Lv and Wang,n.d.). Nowadays, there are a lot of possible practices and tools to address obesity. Due to the proliferation of smartphones and other mobile devices and the wide availability of mobile internet access, many of these tools are now available in the form of smart phone apps or internet platforms. This thesis aims to develop a conversational program (so-called chatbot) that acts as a personalised nutrition coach. It offers its users the possibility to track their daily meals and snacks with an image recognition AI and keep track of their weight. This data is then made available to the user in the form of statistical analysis. Additional features include viewing healthy meal suggestions for all main meals and snacks and setting a goal (maintain or reduce weight), which affects the user’s daily calorie consumption calculated by the chatbot. This approach aims to raise awareness of the user’s daily food consumption. Which in turn should lead to a change in eating behaviour, so-called"mindful eating", which has been proven to lead to healthier eating behaviour according to studies by Jordan, Wang, Donatoni and Meier (2014) and Pintado-Cucarella andRodríguez-Salgado (2016).