Philosophy of the Sciences, 1997
Professor Laurence Goldstein
Lecture 3: Thought Experiments II

1. At the end of the last lecture, I was having an informal discussion with some students about Simon Stevin's thought experiment. A criticism was raised: suppose that one part of the chain were denser than the other (beads more tightly packed on the `necklace'). Would the Stevin reasoning show (wrongly) that that `necklace' would also not move when released ? One answer (I think) is that, after release, the weighted necklace would reach an equilibrium position - with the heavier part at the lowest point - there would not be perpetual motion. But now consider the unweighted necklace. Suppose it slipped to an equilibrium position. But that position would be qualitatively indistinguishable from the starting position (because the chain is homogeneous). Therefore the chain would not move from the starting position. (I've no doubt that this reasoning could be tightened up a bit.)


2. Kathleen Wilkes (Real People, p.6 f.) argues that there is a big difference between thought experiments in science and thought experiments in philosophy. Her reason seems to be that in scientific thought experiments, just as in real scientific experiments, we can hold all variables constant and just manipulate the one that we're interested in. For example, in the Stevin thought experiment, his world of thought is exactly like our world, except that the inclines he envisages are frictionless. Perhaps a more convincing example is of a thought experiment designed to dissuade someone from the belief that equal perimeters must enclose equal areas. Imagine a loop of string 100 cm. long formed into a square. The enclosed area is 625 cm2. Now imagine putting two knitting needles inside the loop and pulling them apart. The resulting shape is 50 cm. long by (say) 1/100 cm. wide, making an enclosed area of 1/2 cm2 - we don't need to measure, we can see `in the mind's eye' that the area is tiny compared with the original square. Here the point is proved without making any idealizations. Compare this with a philosophical example. Wilkes cites Gyges ring which enables a person to disappear at will. The purpose of this thought experiment is to determine whether morality is based entirely on self-interest (for the ring would enable a person to vanish after committing an immoral act, thereby entirely escaping the consequences). Wilkes thinks that such a thought experiment is useless, because the use of this ring brings in its train a whole host of consequences and unanswered questions, so that we no longer can keep our feet on the ground - we are in a world of fantasy where we just can't say what will happen; we can't sensibly draw any conclusions.

3. Wilkes' blanket condemnation of philosophical thought experiments is, I think, too strong. Consider

i. the thought experiment of James Rachels which is designed to show that the American Medical Association is unjustified in forbidding active euthanasia (mercy killing) while allowing passive euthanasia (letting a patient die). What is the moral difference between drowning your 6-year-old cousin and just standing by watching him drown after he has hit his head on the side of the bath, when, with no effort, you could rescue him? No difference -- hence there's no morally relevant difference between killing and letting die. The example is, I think, legitimate and persuasive, even though, ultimately, it may not work.

ii. Searle's `Chinese Room' thought experiment which is supposed to show that no computer, just by running a program, can understand Chinese. Plenty of people find fault with this argument, but it seems wrong to hold that its real fault lies in not being true to life.


4. Thought experiments can be constructive, in that they establish a positive result, or destructive. A destructive thought experiment shows that a conjecture is false by reducing it to absurdity. As an example of a destructive thought experiment, take Galileo's depiction of an inertial frame - this is `destructive' in that its purpose is to destroy the belief that certain contrary-to-fact consequences follow from the supposition that the Earth moves. Galileo invites us not to actually go below deck on a ship and start experimenting with fish in a bowl, a bottle dripping water etc., but just to think about it.

5. Of course, this classification would allow some thought experiments to be both constructive and destructive. If the only alternative to a given conjecture is another interesting and substantial conjecture, then, in doing the destructive work of showing one conjecture false, one is constructively showing the other to be true. For example, if Galileo succeeds in refuting the claim that bodies fall at different rates, he establishes that they fall at the same rate.

6. Back to philosophical question of how thought experiments work, one reaction would be to say that they don't - that they import illicit empirical assumptions. Take Galileo's result about falling bodies, for example. Crucial to the argument is an analogy - the lighter weight would act as a drag on the heavier one just as a slower bicylist would slow you down if he grabbed onto you as you were overtaking him. That's a reasonable assumption, but it's not necessarily true. It's an empirically plausible assumption given what we already know about the behaviour of moving bodies. Likewise with Stevin. We do not know a priori that the string of beads will not rotate ad infinitum. Since the planets are (apparently) in perpetual motion, why not the string of beads? From experience, we know that loss of energy, e.g. through friction, prevents perpetual motion, but Stevin's experiment is supposed to be an idealization - a frictionless plane and no other energy losses. Of course, it seems overwhelmingly plausible that the beads won't move, but that it should seem so is because of what we know empirically about the materials, and we `feel' that, in going from the real cases to ideal ones, nothing is introduced that makes a substantial difference.

7. But aren't there `pure' thought experiments e.g. Maxwell's demon (designed to show that, contrary to classical thermodynamics, there can be instances where there is heat flow from a cold body to a hot one -- the demon lets through the door into the hot chamber only those fast moving molecules from the cold chamber, and the slow molecules in the cold chamber into the hot) and also Huyghens' demonstration that equal elastic masses exchange their velocities on impact ?

8. It is the direct and Platonic thought experiments that are philosophically most problematic, for, in these cases, we appear to be getting `something for nothing', viz., a contingent result about the empirical world but just by using a priori thought. J.R. Brown takes a Platonist line, and argues in Chap.4 of his The Laboratory of the Mind that some of our knowledge of nature is a priori - knowledge of relations between abstract universals, as in the Armstrong-Tooley-Dretske account of the laws of nature. This is a rationalist (as opposed to empiricist) position. It is not one to which many modern scientists would subscribe. One alternative worth considering is that thought-experimental results are just the deductive or inductive consequences of reasonable extrapolations from what we already know empirically to be the case about the actual world. This account contrasts with Mach's (see Sorensen, Thought Experiments, pp.53-4) that we have a purely instinctive cognition that renders superfluous real experimentation with all its flaws.

9. Sorensen considers various kinds of scepticism about thought experiments., e.g., that thought experiments cannot be as informative as real experiments (p.46). For example Franciscus Toletus has a thought experiment to show that vacuums are possible: take a jar full of hot water and cool it to way below freezing point. The contents now occupy less space, so the remainder of the container will be a void. But people contested this conclusion. Obviously, had it been possible then to do the experiment (as opposed to just thinking about it), that would have been far more satisfactory, and would have settled the dispute. Here is the dilemma as Sorensen sets it out on p.48:

10. An answer to this line of thinking, due to Mach, is that what we observe in nature imprints itself surreptitiously on our minds (cf. Nersessian's conception of mental models) so we know instinctively what can't happen (p.51). (So Mach is an empiricist - his position is excellently described in Sorensen, pp.54-55.) If, in the evolution of the human race, knowledge and cognitive predispositions acquired through experience by our animal ancestors gradually became hard-wired, then we might get it all free, a priori and innate. That could explain how, through thought experiments, we are able to obtain synthetic conclusions.

11. Sorensen is sympathetic to much in Mach, but he also thinks that Mach fails to account completely for the fact that thought experiments deliver synthetic (informative) conclusions. For his own account, see Chap.4 etc.

12. For more on the question of why thought experiments work, see D. Gooding, in R. Giere (ed), Cognitive Models of Science. Click for an article on thought experiments from the Stanford Encyclopedia of Philosophy.
GO BACK