From the thought experiment tradition of physics, named by Hans Christian Ørsted (1812) and analyzed by Ernst Mach (Knowledge and Error, 1905), Roy A. Sorensen (Thought Experiments, 1992), and Tamar Szabo Gendler (Thought Experiment, 2000).
Builds a minimal imagined scenario that isolates one question, stripping out confounding detail, so the answer exposes the principle or assumption behind a view.
One error in a board report, varied one feature at a time
A finance director finds an error in a report the day before it goes to the board. Change one feature at a time and watch the intuitive answer: the lines under it name the features the answer ignores and the one it tracks. The opening case has nobody who would ever find out. Illustrative.
Who made the error
Would anyone ever find it
What the error changes
- Change one feature at a time and watch whether the answer moves.Nobody would ever find it, and the answer is still to correct it: the intuition does not track getting caught.Correcting it embarrasses an analyst she depends on, and the answer holds: it does not track who made the error.Shrink the error to a rounding slip and the answer flips, whoever made it and whoever might see it: the intuition tracks what the error changes.
- Change one feature at a time and watch whether the answer moves.Nobody would ever find it, and the answer is still to correct it: the intuition does not track getting caught.Correcting it embarrasses an analyst she depends on, and the answer holds: it does not track who made the error.Shrink the error to a rounding slip and the answer flips, whoever made it and whoever might see it: the intuition tracks what the error changes.
- Nobody would ever find it, and the answer is still to correct it: the intuition does not track getting caught. (known: Never, size: Profit by $400,000)
- Correcting it embarrasses an analyst she depends on, and the answer holds: it does not track who made the error. (maker: Her best analyst, size: Profit by $400,000)
- Shrink the error to a rounding slip and the answer flips, whoever made it and whoever might see it: the intuition tracks what the error changes. (size: A rounding slip)
- Most people's answer
- Pick every rowLog it, fix it next monthCorrect it before it goes
An example
A finance team was writing a rule for errors found in board reports after they had been reviewed. The draft said every error must be escalated and corrected before release. The controllers objected that the rule was really about audit risk, and the analysts said it was really about blame. The argument went in circles because every real example mixed all three.
The finance director put one bare case on the whiteboard: she finds an error in tomorrow's board report. Then she changed one feature at a time. She made it, or her best analyst did. Nobody would ever find it, or the auditors would. It changes profit by $400,000, or it is a rounding slip. The room's answer moved only with the last feature. Who made it and who might see it changed nothing. The rule they adopted tied correction to size: anything over $50,000 or one percent of a line is fixed before release; anything smaller is logged and fixed next month, with no name attached.
Why it works
Real cases arrive with every feature bundled: who did it, who will notice, what it costs, what the relationships are. When people disagree about a real case, they cannot tell which feature is carrying their judgment, so they argue past each other. A thought experiment strips the case to its bones and then varies one feature while holding the rest still, the way a controlled experiment does, using the knowledge people already carry in their heads.
If the answer moves when a feature changes, the judgment depends on it. If the answer stays put, that feature was never doing the work, however much the argument was about it. The method turns a fight about examples into a finding about principles, which a rule can then be written around.
The thinking behind it
Ernst Mach Ernst Mach, the Austrian physicist and philosopher of science, made the thought experiment an object of study in Erkenntnis und Irrtum (Knowledge and Error, 1905). He argued that it draws on experience stored in memory: the experimenter varies the conditions of an imagined case to see what follows, as a preparation for a physical experiment and sometimes as a substitute for one. The term itself, Gedankenexperiment, was first used explicitly by Hans Christian Ørsted in 1812.
Roy A. Sorensen Roy A. Sorensen, a philosopher, treats a thought experiment in Thought Experiments (1992) as an experiment that aims to reach its result without being carried out. Because it is an experiment, it can be judged by the standards that apply to other experiments: whether it isolates the variable it claims to, whether the setup is coherent, whether the result follows. That standard is what keeps a business thought experiment honest.
Tamar Szabo Gendler Tamar Szabo Gendler, a philosopher at Yale who works on imagination and moral psychology and who coined the term alief, examined in Thought Experiment (2000) how imagined cases can teach something new and where they mislead. Her book supplies the caution on this method: an intuition about a stripped-down case is evidence, and an imagined case can still lead the room astray.
How to run it
- Name the question in one line. Write the single judgment in dispute, such as whether an error must be corrected before release.
- Build the barest case. Describe a scenario with only the features people are arguing about, and nothing else.
- List the features to vary. Write each feature with two settings, such as who made the error and how large it is.
- Change one feature at a time. Hold every other feature fixed and ask for the answer again each time.
- Record where the answer moves. Mark the features whose change flips the answer and the ones that change nothing.
- State the principle. Write the rule the answers imply, in terms of the feature that moved them.
- Test it on a real case. Apply the principle to an actual past example and check that the room still agrees.
Where it goes wrong
The first failure is a case too far from life. If the scenario needs a magic button or a person no one would ever be, the intuitions it produces are about fiction. Keep every variant something that could happen next quarter.
The second is changing two features at once. If the size and the audience both change, the answer's move cannot be pinned to either. Write the variants down and check that each differs from the base case in exactly one place.
The third is leading the room. A case phrased to suggest its answer, such as "a small, harmless slip," measures the phrasing. Describe each variant in the same flat words.
When to use something else
- Counterfactual Reasoning. When the question is about a real past event and what would have happened had one thing been different.
- Design of Experiments. When the variants can actually be run and measured.
- First Principles Thinking. When the aim is to rebuild a cost or a design from what is physically or legally required.
Where it sits in SOLVE eX
Go to the source
- Ansicht der chemischen Naturgesetze, 1812
- Knowledge and Error (Erkenntnis und Irrtum), 1905
- Thought Experiments, 1992
- Thought Experiment: On the Powers and Limits of Imaginary Cases, 2000