In this paper we present results on the evolution of informal cooperatives in a population of heterogenous agents. The classical economics approach to cooperative develops models where agents, characterized mainly by their objective success function in a risky activity and their attitude towards risk, have to decide whether to stay in an existing cooperative or leave. After fitting an agent-based model with learning agents on this framework, we expand it to allow agents to create cooperative dynamically, where they decide to associate not only to those whom they think succeed as well as them, but also following other criteria, where friendship and compasion play a role. This paper, mainly exploratory, shows how the definition of network and the circulation of information that takes place in the system impacts on the formation of the cooperative and, as a result, on the distribution of consumption abilities in the population.