Aggression in schools is a problem for which there is no a simple solution. On the other side, it is known that a specific configuration in the distribution of students can affect the behavior among them. Based on a previous experience, we propose to apply genetic algorithms in order to deal with the large number of configurations that can arise on these types of problems. Introducing the concept of penalization has shown to be an interesting concept that allows to reach feasible solutions in reduced computing times. Real environments were considered to conduct the experiments. The set of solutions has been analyzed an accepted as a helpful tool to minimize negative interactions in a classroom. © 2013 Springer-Verlag.