In his comments to my little note on thought experiments, Paul made two challenges. The second of those challenges is not, I think, a real problem. Paul asks whether my criteria for an experiment – that there be a performative stage in it – wouldn’t be fulfilled in going through a syllogism, thus collapsing the distinction between experiment and logic. There are two things to say here. One is that performance in and of itself isn’t sufficient to make an experiment, even though I maintain that it is necessary condition on any experiment. The second is that the material performance of the experiment must be such that it is somehow connected to the design of the experiment. That connection is what the risk in the experiment is all about. So, to use the example of the experiments made on humans in light deprived environments, the performance of the experiment put at risk a hypothesis about the length of human’s circadian rhythm. This is quite different from positing that all men are mortal, and Socrates is a man, or that all dogs are born of bitches, and Lassie is a dog, etc., etc. The inscription of such sentences is not in itself a performance that puts at risk the logical connection between variables and functions one wishes to demonstrate.
It is always possible to transform a thought experiment into some sort of performance. I could easily make a cartoon, for instance, out of Searle’s Chinese Room thought experiment. But the making of a cartoon does not fulfill the performative function of the experiment, which is “about” a man matching Chinese symbols with English symbols.
Paul’s first objection seems similar, but it is, I think, a more potent balestra. He asks whether something like proofs in Euclid’s geometry aren’t performative in just the way I’m maintaining. Whether consciously or unconsciously, Paul is echoing a comment of Lakatos, as cited in this excellent overview of thought-experiments by Michael Stoeltzner at the University of Pittsburg’s Philosophy of Science site.
“Interestingly, Mach held that the purest thought experiments occur in mathematics which, on his account, was economically ordered experience. A similar connection was introduced into modern philosophy of mathematics by Imre Lakatos who contraposed the informal mathematical thought experiment to the formal Euclidean proof. “Thought-experiment (deiknymi) was the most ancient pattern of mathematical proofs.” (1976, p. 9 fn.1) The terminological parallel, to be sure, was drawn by Lakatos because the cited book of Árpád Szabó interprets deiknymi as “to make the truth or falsity of a mathematical statement visible in some way;” (1978, p. 189) with the progress of Greek mathematics deiknymi developed into the technical term for formal proof.”
Why does this differ from the syllogism example? I think the Euclidean proof, viewed in this way, could well be the ancestor of the computer simulation. Computer simulations have an uneasy relationship with the experimental tradition. However, I accept the fact that simulations can produce valid results, can be designed, like experiments, and can have a performative term that encodes risk as in experiments. I accept, in other words, that simulations can be a sub-species of experiment. However, my instinct is that the demonstration is not a computer simulation, because the analogy between design, performance, and result is imperfect.
When I say that the performative term is an encounter with risk, I am merely saying, in my own peculiar way, something that is a commonplace of teaching experimental design. If you look at courses in universities where e.d. is taught, you will notice that it is taught in terms of statistics. This was the great 19th century synthesis. While the popular image is still of the one experiment that proves this or that, the laboratory truth is that the results of iterated experiments vary to some extent, and that experiments themselves are usually grouped together by varying certain elements in the e.d. In other words, instead of a singular phenomenon, the experiment is usually part of a collection of experiments, over which one sums using orthodox statistical methods. I have been trying to avoid highlighting observation, the usual key phrase in talking about experiment, because observation tends to obscure both the performative term of the experiment and the risk inherent to that term – its encoding of such form as would make it statistically available. While computer simulations can be analyzed from this perspective, it is hard to see how Euclidian proofs fit this schema.
To return to the estimable Stoeltzner, he cites some considerable philosophic meandering around this very point by those who make the strong case for thought-experiments as experiments:
“… where are thought experiments located on the scale between theory and experiment? On Norton’s account, they are closer to theory, or at least to the argumentative analysis of an experiment, and they can accommodate rather general philosophical principles into a scientific argument. Andrew D. Irvine holds that “the parallel between physical experiments and thought experiments is a strong one.” (1991, p. 150) All assumptions of a thought experiment must be supported by independently confirmed observations and it typically has repercussions on a certain background theory. On Irvine’s account, the fact that “many thought experiments are meant to precede real experiments in which the original thought experiment’s premises are actually instantiated” (Ibid., p. 151) and the fact that some elements of a thought experiment are assumed to be true, proves that it typically contains some but not only counterfactual elements. Ronald Laymon proposes to render benign the counterfactual character of thought experiments involving frictionless surfaces and the like by treating them as “ideal limits of real experimentation.” (1991, p. 167)”
LI’s case against the thought experiment as experiment depends, in part, on showing that thought experiments aren’t very much like experiments. But we are not experts in experimentation. We’ve pointed to obvious problems that would occur to anyone. We are, however, sensitive to status cues in texts, and we are going to turn, in our next post on this topic, to the prestige of the experiment. Our hypothesis is that the pullulation of thought experiments in philosophy has less to do with the epistemological advantages of experimentation and more to do with the prestige accrued by the image of the experiment – what Barthes would call the myth of the experiment.