The “both … and …” view

De Mey, T. (2003). The dual nature view of thought experiments. Philosophica, 72, 61-68

What De Mey tries in this article is to show that the argument view and the experiment view on thought experiments are just that, namely, views. Neither of them is true – in the sense of capturing all there is to capture about thought experiments – but neither of them is false – in the sense of being logically rejected by the others. Thought experiments are both.

According to contemporary wisdom, one either holds that thought experiments are, basically, experiments or one subscribes to the rather deflationary view that they are, deep down, arguments 62

Now, De Mey’s plan is methodologically very shipshape: let’s first ask ourselves why we are wrestling with thought experiments in the first place, and then see what is the “nature” of thought experiments. Insofar as the theoretical (or meta-theoretical) purpose is in the blur, talk of whose definition is better is blank.

There are three problems, De Mey recognizes: (1) the problem of source of knowledge, (2) the problem of heuristic value, (3) the problem of evidential significance.

First: where do the goodies come from? “The very possibility of acquiring knowledge by means of thought experiments is generally taken to be a problem for empiricists” since it makes this knowledge difficult to trace back to experience. One way out of this is Norton’s empiricist account:

Thought experiments in physics provide or purport to provide us information about the physical world. Since they are thought experiments rather thanphysicaZ experiments, this information does not come from the reporting of new empirical data. Thus there is only one non-controversial source from which this information can come: it is elicited from information we already have by an identifiable argument, although that argument might not be laid out in detail in the statement of the thought experiment. The alternative to this view is to suppose that thought experiments provide some new and even mysterious route to knowledge of the physical world.

The “mysterious route” being surely Brown’s platonic account. Some thought experiments, Brown argues, do two things: they destroy a theory and build a new one. While the first part might be reconstructible as argumentation, the second part is not. Also, the second part does not involve new knowledge, nor does it derive its claim from previous data. (This should be taken with a grain of salt, since the fact that some conjunction of hypotheses cannot be the case can already be qualified as knowledge.)

Now, in connection to the second problem (the “heuristic value”), De Mey adds: “Whatever the merits of Norton's argument view are, it does not explain why thought experiments can be "psychologically helpful" (and thereby rhetorically effective)” (65). I agree with De Mey, but I think we hold this view for different reasons. He construes Norton’s account as theoretically incapable of shedding light on the rhetorical prowess of TE’s, while I construe Norton’s account as theoretically undeveloped, with the same result. In other words, is not like the account cannot say something explicatory about the heuristic value, it is that it does not. Just one step forward: although it does not, not only that it can, but it seems to me like it is the only one that can give a satisfactory answer. And half the road is already travelled: they are rhetorically effective because they are good arguments!

As for their epistemic significance, De Mey (2003, p. 66) speaks of the role TE's play in theory choice as their most "spectacular capacity" - whereas I think one should construe it as one of their defining ones. I think we would miss the property of terms if we would speak of a thought experiment every time someone supposes something. If any supposition triggers a thought experiment, then the label loses its classificatory values, i.e. it ceases to isolate a group of entities for the purposes it used to.

Norton's empiricist solution to the problem of the source of thought experimental knowledge can give us a first hint. Nobody doubts that arguments can play a role in theory choice. So, as far as thought experiments are arguments, their evidential significance seems fairly unproblematic (66)

There’s also Bishop (1998; 1999). He claims that the clock-in-the-box event works as an illustration as to why thought experiments are not arguments. The clock-in-the-box was a failed attempt by Einstein to confront Bohr with a counterexample to the uncertainty principle. It was considered a failure because, using the same set-up, Bohr pointed out that Einstein was missing something: in order to weigh the box, it must move in a gravitational field - since it does, the uncertainty principle (the measurements in question being mass and energy) holds.

Ok, now De Mey’s view. While beginning, he uses words in a certain way and, although he slides quickly to the more urgent problem, I believe he makes a good choice which needs underlining. Notice that for him, the thought experiment is not an experiment and an argument but a description of both:

thought experiments like that of the clock-in-the-box have a dual structure: they involve (1) the description of an imaginary situation and (2) the description of its settlement or winding up.

Their “evidential significance” is different according to what points (1) and/or (2) we choose to take into account. So much so that De Mey speaks of evidential significance₁ and evidential significance₂.

The last step for De Mey is to take a modern account of experimentation and cast thought experiments into it.

Sophisticated conceptions of experimentation, typically invoke more structural dimensions. Radder (1996), e. g., differentiates between three "phases" of experimentation: (1) preparation, (2) interaction and (3) detection. Firstly, during preparation the object and the apparatus are prepared in agreement with the plan of the experiment. Subsequently, interaction results in the transfer of information from the object to the apparatus. Finally, detection involves obtaining the information "by measuring or observing the relevant property of the apparatus (Radder 1996: 11).

Each phase (“make the device”, “use the device”, “read the device”) has a material realization and a theoretical interpretation. The first one is (or should be) theoretically-independent. The second one is theory-driven.

So there you have the solution:

So there are two ways to describe a thought-experimental process, i.e. in terms of its material realization and in terms of its interpretation. To do full justice to the evidential significance of a thought experiment then, of whatever kind it is, we need both. As far as the material realization of the thought-experimental process is what adherents of the experiment view on the nature of thought experiments stress (as I believe they do) and as far as the theoretical interpretation of the thought-experimental process is what adherents of the argument view have in mind (as I believe they do), we can safely conclude that it doesn't make sense to say that thought experiments are, basically, either experiments are arguments. They are, deep down, both experiments and arguments 76