One idea I had was to have students write down the instructions they gave to ChatGPT, then the ChatGPT output and then critique the ChatGPT output. This raises the question: Can ChatGPT “critique” its own output as a second round thing?