Modern teaching tools, such as educational robotics, require new learning environments. The teacher especially needs to be supported in novel ways. Conflative learning environment is based on the agents collecting data from the learning process. The teacher can build a support environment based on his or her empirical observations from the classroom. An implementation of the conflative learning environment is described and two case studies about the use of system are reported. The results show that agents are useful and efficient in data collection, and that the Empirical Modelling environment can be used to construct the working classroom models.