Peer effects are a subject of increasing attention in many areas of economic research. Peer influence can create social multiplier effects, whereby an initial investment targeting one small group can lead to larger changes, as individuals close to the target group are directly influenced by its actions. When it comes to the green transition, a central question is that of learning about green agricultural technologies. The adoption of a new technique sometimes requires specific technical assistance, which cannot easily be provided to all eligible farmers because of the high costs it would entail. In this context, observational learning can play a crucial role in the diffusion of new practices. However, there are many reasons why social learning might not ultimately happen; and it is often difficult to identify and measure it accurately. The literature provides several examples of the diffusion of agricultural innovations through social networks and peer effects in developing countries. There is, however, no prior evidence of the existence of the diffusion of agroecological practices through social learning in developed countries. We aim at filling this gap by studying the knowledge spillovers of an intervention that trains a number of volunteer farmers in new agricultural practices, with the aim of indirectly transforming those of the members of their network – we call a network of farmers a cluster. The experiment requires the implementation of a pilot program, in which the clusters are randomly assigned to training. The proportion of farmers within each cluster who are offered training varies between treatment arms. We call them the injection points of new knowledge. This design allows first to infer the causal effect of training on treated groups, then to check the presence of the diffusion of new knowledge inside the clusters, and finally to highlight a threshold in the proportion of injection points necessary for the diffusion of new knowledge.