Connecting physical activity with context and motivation: a user study to define variables to integrate into mobile health recommenders

Abstract

In this paper, we aim to improve existing health recommender systems by defining relevant contextual and motivational variables to recommend physical activities and collect appreciation feedback. Current health recommenders do not sufficiently include users' context and motivational theory when personalizing health suggestions. To bridge these gaps, we conducted a 21-day longitudinal user study with 36 participants using our Android app with collected sensor data and Ecological Momentary Assessments to collect daily activities, mood, and motivation. This study resulted in a dataset of 724 activities. Two approaches to determine feature relevance were followed: variable importances analysis on 40 input variables, and statistical analysis of mean differences in outcome variables across contexts. Our findings suggest recommending activity duration, intensity, location, and type by incorporating: company, situation (e.g., free time or work), happiness, calmness, energy level, physical complaints, and motivation. As such, we propose opportunities for future health recommenders to integrate these data with contextual pre-filtering techniques, extended with our suggestions for automatically collected weather, location types, step count, and time. We also propose to use mood and motivation as appreciation feedback to focus on user well-being and boost motivation.