Assign a subject to a fairly long test, at least an hour or more, where they are rewarded for correct answers and harshly scolded/yelled at/possibly even physically hurt like a small shock or loud buzzer, for wrong answers.
Except that's just what they're told, while the actual criteria for being rewarded or punished is literally decided by RNG. It still keeps up the facade that their input matters and that they can improve their outcomes by learning and doing better, but that never is reflected in their results because their actions do not matter. And when they do get a "wrong answer", they're never told what the "correct answer" was. Just blindly being punished.
To make it worse, have the questions be a mix of general, easy questions, more niche trivia ones, and very complex specialized questions. So they both get the unease of the unfamiliar subject matters, while feeling confident about the common sense ones, and both of those get flipped on their heads when their blind guesses of hard questions seem to keep being correct while their confident answers of easy questions somehow keeps being called wrong.
It would be a very cruel experiment into the concept of learned helplessness, especially when combined with the lie of being meritocratic. "If you just made better choices and were less ignorant, you would be happier!" When the researchers know for a fact nothing the subject does can change what happens. All the effect their input has is, if they freeze up or get too scared to keep answering, the entire ordeal will just take even longer before they're freed from it.
