User Modeling for Physical Activity Tracking

Project description

Physical activity tracking wearables and apps constitute an environment where user input (set goals, demographics, preferences) and sensed input (steps, calories burnt, heart rate, weather) provide good opportunities to explore user modelling techniques (ie computational representations of users that can be used for personalisation or recommendation purposes). As a result, one could explore factors of interest and test hypotheses (eg "are people more active on sunny days?").

Typically, digital phenotyping studies focus on monitoring health conditions that are chronic, episodic or seasonal, but little attention has been paid to understanding the association of human behaviour and wellness -- although there are a few exceptions, see [1].

This project is about building user models using sensed data and physical activity data in order to explore what are the factors that facilitate/prevent achieving physical activity goals and intervene accordingly. To do so, the PhD project may follow this plan:

1. A review of the literature review and formulation of the research hypotheses
2. Build a physical activity tracking systems using the AWARE framework and its Fitbit plugin.
3. Run longitudinal user studies to collect data.
4. Identify the factors that prevent/facilitate goal achievement and build user models accordingly.
5. Based on personalisation principles (see an extensive review in [2]) and informed by user models, design and evaluate interventions to support users in achieving their physical activity goals.

Consequently, the project would be appropriate for a candidate who is skilled in at least two of the following: software development, experimental design and data science. If one skill was missing, the candidate must show commitment to acquire it during the course of the project. Also, it is expected that the candidate will publish academic papers in SIGCHI venues (CHI, DIS, EICS), UMAP and UbiComp/IMWUT.