People are social beings in nature and our friends and family play a crucial part in our lives. In fact, many research has shown that maintaining meaningful relationships has health benefits, such as lower risk of depression and heart disease and high probability of longevity. However, existing work has not understood what specific activities we do with our friends and family that give us the health benefits. Part of the constraint is the impossible task of collecting continuous data on all activities people do throughout the day. With the popularity of mobile phones, more and more people carry their personal phones with them 24/7, which contain powerful sensors, such as GPS and accelerometers. These sensors can collect valuable interactions the owners have. This research examines what daily behaviors, collected from sensors on mobile phones, people do with their friends and family members that affect their mental health. Using machine learning algorithms and statistical models, the project aims to clarify the associations between social behaviors and mental health.