College Should Be Way More Fun

One afternoon last fall, a class full of Amherst seniors forgot I was there. In the 19th-century octagonal room where I taught my course on fiction, they were deep in an argument about the tempestuous ending of Henry James’s The Turn of the Screw—about whether the ghosts haunting two children in a gothic country house are real, about whether they exist only in the deteriorating mind of their governess, about why one of the children dies at the novel’s conclusion, about whether he even dies at all. The famously ambiguous novel is strewn with evidence to support incompatible interpretations, and my students found it all. The discussion became loud, animated. People smiled, then laughed. Nobody was waiting for me to tell them the answer; the room was theirs, all eight sides of it.A large language model on one of their phones would have exhausted the debate with just a few keystrokes. Try it: Ask ChatGPT or Gemini if the ghosts in The Turn of the Screw are real, and they will with alarming speed give you a few bullet points for rival interpretations—and then stand ready for the next question. Ask one to pick a side, and it will do so with triumphant certainty. (“Definitively? No—the ghosts do not exist,” ChatGPT told me.) Or it might offer you a cheeky riff to tie things off, as Claude recently did for me: “The ‘real’ answer may simply be that James wanted the question to haunt you.” The ghosts are haunting, get it?My point is not that the LLMs are more right or wrong than their human counterparts, but that the speed at which they churn through the argument is the exact opposite of the slow, messy conversation that unfolded in front of me last fall. What makes The Turn of the Screw so generative isn’t that it has a hidden answer waiting to be unlocked. James built the ambiguity in on purpose, and lingering over that uncertainty, turning it over, is the entire point. (“The story,” as one of the characters famously says, “won’t tell.”) That kind of intellectual experience