What Is a Thought Experiment? Reasoning That Never Needs a Laboratory

A thought experiment sets up a hypothetical scenario precisely enough that its consequences can be reasoned through, and uses the result to support or destroy a general claim. The commonest destructive use is to construct a case where an accepted principle produces an absurd or contradictory result, which refutes the principle without requiring any observation. Galileo's argument against the Aristotelian claim that heavier bodies fall faster works exactly this way: imagine a heavy body tied to a light one, and the combination should fall faster than the heavy body alone because it is heavier, and also slower because the light body retards the heavy one, which is a contradiction, so the original claim cannot be right. No tower and no dropping is needed, and the argument is conclusive given its premises. Constructive uses build intuition about a situation that cannot be observed, which is how much of relativity was developed, and elicitive uses probe what a concept actually requires by testing it against unusual cases.

A handful have been consequential enough to shape whole fields:

• •Einstein imagining riding alongside a light beam, and later the equivalence of a person in an accelerating box and one in a gravitational field, which became the founding insight of general relativity
• •Maxwell's demon, a creature sorting fast and slow molecules to violate the second law of thermodynamics, which took over a century to resolve and led to a deep connection between information and entropy
• •Schrodinger's cat, devised to show that the quantum formalism produces absurd results when scaled up, which its author intended as a criticism and which is now most often cited as a description
• •The trolley problem and its variants, which separate different moral principles by constructing cases where they disagree
• •Searle's Chinese room, arguing that manipulating symbols according to rules cannot constitute understanding, which remains the standard reference point in arguments about machine intelligence
• •Rawls's veil of ignorance, asking what principles of justice people would choose without knowing their own position in society

The obvious objection is that no new information about the world enters the reasoning, so nothing about the world can come out, and several answers are given. One holds that the method simply makes explicit what was already implicit in beliefs the reasoner held, so the thought experiment reorganises existing knowledge rather than adding to it, which explains how it can be informative without being empirical. Another holds that reasoners run a mental model that draws on deep familiarity with how things behave, so the exercise is a form of experiment conducted on internalised physics, which explains why such arguments sometimes succeed and sometimes fail badly. A third, more deflationary, treats them as arguments in narrative dress, whose persuasive force comes from vivid presentation rather than from any distinct method, and which can therefore be restated as ordinary deductive arguments without loss. The disagreement is live, and the practical relevance is that vividness is a known hazard, since a compelling story can carry an argument whose premises would not survive plain statement.

The failure modes are well catalogued. Smuggled premises are the commonest, where the scenario quietly assumes what it purports to demonstrate. Impossible scenarios undermine the conclusion, since if the situation could not occur under the laws being discussed, reasoning about it proves nothing about those laws, and a good deal of debate about specific cases turns on whether the scenario is genuinely coherent. Intuitions may simply be unreliable guides, and experimental philosophy has tested reactions to classic cases across populations and found substantial variation by culture, by the order in which cases are presented and by irrelevant framing, which undercuts arguments resting on the claim that everyone would respond a particular way. Over-idealisation removes exactly the features that make the real case hard. And the trolley problem has attracted specific criticism from within ethics on the grounds that repeated exposure to artificial dilemmas trains a way of thinking about moral questions that transfers badly to actual decisions, where the options are not fixed and the facts are not given.

Setting up a hypothetical case precisely enough to reason through it can refute a general principle by deriving a contradiction, which is how Galileo destroyed the claim that heavier bodies fall faster without dropping anything. Einstein's accelerating box founded general relativity. The puzzle is how reasoning with no new information teaches anything, and the failure modes are smuggled premises, incoherent scenarios and unreliable intuitions.