Social interventions are a popular tool to stimulate pro-social (including climate-friendly) behavior. Their use is, however, limited when the descriptive norm is low, i.e. when a desirable behavior is only practiced by a minority within the respective reference group. We tackle this issue by testing new strategies for social interventions, with an especially sophisticated target group. We implement a field experiment at two subsequent conferences in environmental economics, with which we examine the conference participants' proclivity to offset carbon emissions. For the two treatment conditions that we introduce, we document an average null effect. Yet, for one condition, we find that the intervention can be effective when the targeted individuals feel socially close to the referenced peer group. Further, we find suggestive evidence that the effectiveness of such interventions increases as individuals are exposed to repeated treatment, but with decreasing marginal returns.