A new kind of supervised learning approach is proposed to estimate the number of occupants in a room. It introduces the concept of interactive learning where actual occupancy is interactively requested to occupants when it is the most valuable to limit the number of interactions. Occupancy estimation algorithms rely on machine learning: they use information collected from occupants together with common sensors measuring motion detection , power consumption or CO 2 concentration for instance. Two different classifiers are considered for occupancy estimation with interactions: a decision tree C4.5 classifier and parameterized rule based classifier. In this paper, the question of when asking to occupants is investigated. This approach avoids the usage of a camera the determine the actual occupancy.