Motivation

Problem Formulation

Implementation
System Overview
Hardware
Interaction Dialog

Discussion

Motivation

The motivation behind this project is quite simple. I am interested in proactive healthcare and specifically motivating people who lead a sedentary life style to undertake moderate exercise such as taking a leisurely walk. I would like to use behavioral change methods, such as the transtheoretical model, and other social psychology to create intrinsic motivation (motivation that is internally motivated rather than motivation from an external actor). Specific to this class,  it seemed as though people would be more responsive to a suggestion such as "Why don't you take a walk" based on their current emotion. For example, someone may be more willing to take a walk if they are incredibly stressed - as a stress relief - as opposed to someone else who may be happy and just want to go enjoy the world. In any case, if emotion can be sensed the time of motivation, what behavior is being motivated, and the method or "tone" of the motivation (such as pleading versus authoritative) can be modulated to be most effective in bring about an intrinsic and lest burdensome to the user. 

Further, there are many studies that show that emotions such as stress can have serious health consequences. By understanding changes of emotions and alerting the user to these swings may help the user become more "even keeled"

Even though the above formulation seems concrete, there are many issues underlying the development of such a system. One of the core aspects of such a system is knowing when to interrupt the user with a exercise based motivation. For this class I decided to focus on the problem of how to use emotion for interruption. My intitial direction was motivated by recent work done by Liu & Picard on using heart rate related to stress and empathy to reduce the burden of interruptions by trying to find more 'interruptible' times. This work gave me a basis for my measure of affect - heart rate. Another recent piece of Media Lab work by Ho & Intille which used accelerometers to understand 'transitional states' and interrupted the user based on these states. The work by Ho & Intille nicely complements the work by Liu & Picard since one problem in the later work is the fact that heart rate changes with physical activity. Also, knowing when the user is getting physical activity is important in for motivating users who lead a sedentary life. Thus, knowing the state of the user (such as walking, sitting, standing) is quite helpful for informing the interruption and for distinguishing between hear rate change due to physical activity versus that due to emotional change. Further, work by Turnball (2002) showed evidence that the arousal due to a bout of exercise can be misattributed to another event if that event occurs around 10 minutes after the exercise bout. It is important that a behavioral  device is aware that this may happen and help the user understand why they feel good - due to exercise and not eating the ice-cream after their exercise. It is thus desirable for multiple reasons to know both physical activity and emotional state and how they interact.