There is a complex unwritten code which regulates human interactions. In this paper we present a camera based mon- itoring system that explores the relationship between prox- emics, visual attention and personality traits during interac- tion. People’s relative positions an head poses are extracted with a multi-target tracking algorithm and used to (i) esti- mate, under the ’thin slices’ hypothesis, the level of extro- version and neuroticism of a person, to (ii) learn a model of people interactive behavior that improves, once integrated in the algorithm, the accuracy of the tracking estimates, and to (iii) steer a set of active cameras to the subject found to exhibit the most peculiar interaction pattern. We report on experimental results in a natural scenario where people engage in a party.