Very few learning systems applied to problem solving have focused on learning operator definitions from the interaction with a completely unknown environment. Autonomous Intelligent Systems (AIS) deal with that issue by means of architectures where learning is achieved by establishing plans, executing those plans in the environment, analyzing the results of the execution, and combining new evidence with prior evidence. This paper proposes a selective mechanism of learning allowing an AIS to learn new operators by receiving them from another AIS in a higher stage in the Learning Life Cycle (LLC) with more cycles of interaction in the environment. The proposed collaboration mechanism also considers how to deal with theory ponderation (operators ponderation) and how to include the new operators (provided for) in the set of theories of the receiver AIS. The experimental results show how using collaboration-based learning among AIS provides a better percentage of successful plans, plus an improved convergence rate, than the individual AIS alone.