Philosophers routinely argue from cases that never happened and often could not happen. The device is older than the discipline's modern form and does specific work that examples from experience cannot.
They isolate one variable
Real situations arrive with everything attached at once, which makes it difficult to tell which feature is doing the moral or conceptual work.
An invented case can hold everything constant and change one thing, in the way an experiment does, so that a reader's shift in judgement can be attributed to that change.
The artificiality that critics object to is therefore the point. A scenario that resembled ordinary life closely would reintroduce all the confounding detail the case was built to strip out.
They test principles at the edges
A principle that works comfortably in familiar cases may produce absurd results in extreme ones, and finding that boundary is informative.
The usual pattern presents a case where a stated rule gives an answer almost nobody accepts, which forces either a revision of the rule or an explanation of why the case is special.
Neither outcome is a defeat. Being made to state the exception explicitly is progress, because an unstated exception is a hidden commitment that has never been examined.
Religious traditions used them long before philosophy departments
Hypothetical cases appear throughout legal and ethical literature in several traditions, where jurists reason about situations that had not arisen in order to prepare for those that might.
Some of these scenarios are deliberately extreme, and commentators sometimes note that the case is unlikely while still working through it carefully.
The practice trains judgement. A community that has reasoned about difficult imagined cases has vocabulary ready when a real one appears under time pressure.
Construction determines what a case can show
A scenario can be built to make one answer feel obvious, through the details chosen, the order of presentation and the words used to describe the people in it.
Small changes in wording produce measurably different responses, which is why the same underlying dilemma can be presented in ways that reverse the intuition it elicits.
Careful users therefore present variants rather than a single version, and treat a judgement that survives rewording as more informative than one that does not.
Intuitions are evidence, not verdicts
What a thought experiment produces is a reaction, and the argument only begins once that reaction is on the table.
Reactions vary between people and between cultures, and a case that seems decisive to one group can leave another unmoved for reasons worth investigating in themselves.
Treating an intuition as data to be explained, rather than as a conclusion already reached, is what separates the technique from a rhetorical trick.