**1. Introduction**

Physical inactivity is an increasingly serious health problem: the World Health Organization (WHO) has identified that it is the fourth leading risk factor for global mortality [1]. The organization estimates that a lack of physical activity leads to 3.2 million deaths per year globally. Physical inactivity has all to do with modern sedentary lifestyles, which are led by 60% to 85% of people worldwide, according to the WHO. One aspect of a sedentary lifestyle is that people are more inclined to passive modes of transportation. Active travelling modes such as biking and walking can contribute to a healthy level of physical activity [2]. Another aspect of a sedentary lifestyle is related to the work environment, where much work is done by people seated in chairs in front of computers. Research suggests that having desk jobs increases health risks up to 50%. Integrating small activities in work routines can help to increase physical activity and lower health risks [3].

At the same time, modern mobile technology, such as smartphones and digital measurement devices, provides new opportunities to tackle physical inactivity. In 2015, 43% of the adults worldwide owned a smartphone, with percentages up to 70% for developed countries [4]. Smartphones allow for continuous and real-time monitoring of activity behavior via built-in sensors such as accelerometers, and provide possibilities for giving contextualized and personalized feedback. This makes the smartphone a potentially powerful device for real-time coaching of people towards a more active lifestyle. To use smartphones and sensors for this aim, they should be integrated into a behavior change support system, which is defined as an information system designed to form, alter, or reinforce attitudes or behaviors [5].

In this paper, we describe the design of such a system in detail, together with lessons learnt and suggestions for future developments. The system is developed in context of an interdisciplinary research project and is called Active2Gether. The goal of the project is to combine domain knowledge from experts in physical activity interventions with modern mobile technology to design

an intervention that encourages physical activity among healthy young adults. One of the innovative aspects of the system is that it exploits model-based reasoning techniques for tailoring the coaching to the needs of the user. Up to now, this has hardly been applied within existing interventions [6].

As the name suggests, social processes play an important role in the Active2Gether system. This is reflected in different ways, such as the implementation of social comparison mechanisms on both an individual and a group level. In addition, the system addresses psychological constructs as social norms and social aspects of outcome expectations in its coaching messages.

The aim of the Active2Gether system is to increase or maintain levels of physical activity among young adults in the age group of 18 to 30 years. The system is being evaluated in a trial (see [7] for a detailed description) in which over 100 participants, aged between 18 and 30 years old, used either a variant of the Active2Gether system or the standard website that belongs to a commercial activity tracker for approximately three months. The user evaluation of the system by the participants is described in [8]; in this paper, we focus on the architecture and functionality of the system.
