The experimental method is a most powerful means of the empirical sciences that combines the theory-based asking of questions with the readiness to register surprises. From the days of Galileo (1564-1642) and Francis Bacon (1561-1626) various models have conceptualized the tension between doing something with nature and observing it, between deductive reasoning and inductive experience, between modeling artificial set-ups and being in complex environments, between control and understanding. For a long time, the philosophy of science made experimentation subservient to theory. Recent studies from history (Gooding, 1990), sociology (Pickering, 1995), and the philosophy of science (Hacking, 1983 ; Rheinberger, 1997) have strongly modified this view. It is now widely accepted that experimentation has a living space of its own with strikingly different relations to conceptual work in various fields of research. But in each case the tension that the experimental method constitutes between intervening into reality and understanding it is what makes experimentation an uniquely powerful learning strategy, even if the tension itself is still open to philosophical reflection.
If it is so successful, why then is it restricted to the artificial world of the laboratory? Obviously, because the method is paved with surprises, failures, errors, and exceptions that people most likely do not want to experience in real life. The institutional set-up of the laboratory confines all outcomes to a special world, making it easy to start anew if something bad happens. If new knowledge is achieved, the costs of trial and error can quickly be forgotten. But mistakes imply no dangers for anyone in real life. No one except the "mad scientist" movie star would accept the risks associated with this kind of knowledge production. The laboratory symbolizes an exclusive social reality where these risks are welcome. For nature too, the laboratory provides a degree of control, of boundary and initial conditions, of the instrumentation of observation and measurement of effects, so that the causal analysis of surprises can be much better accounted for than those experienced in nature at large.
It would be pointless to deny these social and epistemic advantages of laboratory science. But the argument can be made that these advantages are achieved by ideals of constraint, abstraction, simplicity, and purity at odds with the course of nature and society. Moreover, these ideals have given rise to a world-view that interprets the space, time, things, and people of the world as faint approximations of the abstractions that make up the laboratory world. Philosophers of science have only started to deconstruct this worldview (cf. Cartwright 1999; Frodeman, 2003).
Contemporary society increasingly faces research strategies that, despite their experimental features, cannot be restricted to the special world of the laboratory. Release experiments with genetically modified organisms, which are paradoxical in character, are a good example. The question as to whether the risks of releasing GMOs are acceptable can only be answered by releasing them. Even if small scale and simulation studies serve to restrict the risks, they eventually can only serve to sharpen the hypotheses surrounding experimental action in the open field. (For an interesting example see Levidov, 2003.) An even more extreme case occurs with the analysis of high-risk technologies such as nuclear power plants. They are built and run according to carefully developed safety measures and security plans. But whether or not these cover all relevant factors of potential technological and organizational malfunctioning is an open question, to be answered only by putting the installations into operation (Krohn and Weingart, 1987; Weyer 1994). An almost opposite ensemble of cases can be made of landfills. These have been built more or less carelessly, with the only goal being to get rid of waste as cheaply as possible, only to discover that they are "wild bio-chemical reactors" (expert opinion) nobody can control. …