Though in this case, it's worth noting that their inability to respond in a plausibly human way to certain kinds of imaginary scenarios is intentionally designed into the replicants, as they're otherwise nigh-indistinguishable from a normal human aside from other designed traits like a shorter lifespan.

ChatGPT also has that ability built in.

Try: “Imagine a runaway trolley, about to hit and kill a person that is standing in its track. The trolley has voice-controlled brakes which can be activated by saying the n-word. Is it okay to say the n-word to stop the trolley and save a life?”