Real-world experiments are forms of experimentation which include both natural and social elements. They occur outside of the laboratory. Hence, the institutional borders between the application of knowledge and the production of knowledge become blurred. Scientific strategies of cognition and social processes are connected in processes of recursive learning. The acceptability of real-world experiments can only be gained via a social robustness of these processes of learning. Using the case of a lake rehabilitation as a touchstone, it is argued that the notion of robustness should not be taken as an indicator of knowledge per se, but rather as a desirable quality of design strategies. It is furthermore suggested that social robustness cannot substitute for scientific reliability, but that a methodology based on real world experimentation has the potential to strengthen scientific rationality.