We analyze the users' intimacy to investigate the differences in smartphone usage, considering the user's location and number and kind of people physically around the user. With a first user study we (1) validate the intimacy concept, (2) evaluate its correlation to smartphone usage features and (3) we computationally model it. Shorter, more frequent, and less engaging interactions take place when intimacy is lower, while longer, less frequent, and engaging interactions when intimacy is higher. With a second user study, we investigate the intimacy predictability in practice. Location–time features are predictive for the intimacy, and other smartphone-based features can improve the intimacy prediction accuracy.