Mechanism design is applied science carried out under the most favorable conditions – in conditions that maximize the chance of successfully exporting experimental results outside laboratory walls. In this chapter, I try to generalize from this particular case to provide an account of external validity inferences in less than ideal circumstances. Scientists, as a matter of fact, do draw inferences from experiments, even when the outside world cannot be shaped to resemble laboratory conditions. I introduce a new example from auction theory in which the procedure is exactly symmetrical to the one followed by the FCC consultants: the laboratory conditions were shaped so as to resemble those of a real-world economy, instead of the other way around. I also draw some analogies from experimental medicine and elaborate on the role of laboratory and field evidence in external validity inferences. Before doing that, however, I have to dispose of a disturbing and extreme stance on the external validity problem, which I label “radical localism.”

Radical localism

The cathodic rays that make our TV sets function are generated by a carefully engineered causal structure that has been repeatedly tested in laboratory conditions, then in the factory, and finally stabilized, “shielded” within a plastic box, and sold to customers all over the world. With few exceptions (e.g., Cartwright 1999), philosophers of science have neglected this pervasive feature of applied science (or “techno-science,” as it is sometimes called): success is usually achieved as much by construction as by accurate representation.