Tools for Evaluating the Consequences of Prior Knowledge, but no Experiments. On the Role of Computer Simulations in Science
|Table of Contents|
|2 Common features of simulations and experiments|
|3 Distinguishing features of experiments and simulations|
|4 Borderline cases|
|4.1 Experimentum crucis and analog simulations|
|4.2 Simulation-like experiments|
|4.3 Experiment-like computer simulations|
|4.4 Hybrid simulation-experiments|
|5 Summary and conclusion: Computer simulations as a tools for drawing conclusions from prior knowledge|
The last example gives rise to the notion of simulation-like experiments. Simulation-like experiments can be understood as experiments where none of the above mentioned distinguishing features becomes relevant. These are experiments that do not operate directly on the target system but on an object representing the target system, in which case also the empirical data they provide is not data about the target system but about the object that is used in the experiment to represent the target system. Or, if they do operate on the target system, it is merely for convenience and the data they provide is not expected to be any different from that which would have been predicted by models of the target system based on known theories. And finally, simulation-like experiments can only be experiments which are not employed to test fundamental hypotheses.
The distinction between simulation-like experiments and other types of experiments may not be very strict. This holds in particular if it is made a requirement that simulation-like experiments do not operate on the target system directly but, like computer simulations, substitute the target sys•tem with a representation of it. Does a scale-model, for example, operate on the target system or not? It could be argued yes, because the scale-model is of the “same stuff” and the same natural laws take effect on it. But it could also argued no, because a scaled down model is not the real thing itself.
On the other side, the distinction between simulation-like experiments and computer simulations appears to be reasonably clear. Computer simulations operate on a symbolic representation of their target system, but the representation of the target system in simulation-like experiments is not symbolic. Thus, simulation-like experiments and computer simulations can be distinguished, although this difference may not be epistemically relevant in the above defined sense. For, if the symbolic representation in the computer is good enough, computer simulations can replace simulation-like experiments.
It is important to keep in mind that even though some experiments are simulation-like and therefore not easily distinguished from simulations, it would be wrong to take this as an indication that simulations in turn are generally like experiments.