Cherenzoo is an educational project for automatic routines that will make use of the results of perception experiments conducted on humans through structured interviews. The goal is to use pareidolia to draw up a database of solutions to some pattern recognition problems still unsolved, and specifically, for the classification of the traces of secondary cosmic rays captured by Cherenkov telescopes. The database could be then used to train Machine learning algorithm. To this aim, we developed an activity built for a wide range of users’ ages with a specific design focused for children of kindergarten and the first two primary school classes, to take advantage of their lack of preconceived ideas. Hence, the proposed activity has a dual character: the part that requires the involvement of children is a narrative game which obviously does not include the previous explanation of the complex astrophysical problem underlying the images to be catalogued. On the other hand, adults and young adults, after having learned as much as possible about the astrophysical problem, can sort by eye the cosmic ray traces. All the classifications, those done by children as well as those coming from the adults cataloguing game, will serve to build a data base of training cases for automatic algorithms, with the aim of improving their ability of discerning ambiguous traces of particles and photons.