Can AI help students think harder?

Chen Chih-nung (陳啟濃), a former principal at Kuohsin Junior High School in Nantou County, wrote about writer Chu Yu-hsun’s (朱宥勳) frustration with artificial intelligence (AI) in creative writing (“Students losing their minds over AI,” Sept. 30, page 8). Chu said that he would refuse to judge any literary award that did not explicitly prohibit the use of AI.

A good story told powerfully and beautifully, one that resonates with readers and touches their hearts, is a good story regardless of how it is generated.

Literary competitions have always had to establish authorship. A 30-minute interview with shortlisted writers could provide conclusive evidence to show whether a piece was written by AI, with the help of technology or entirely by the author.

Writers know their work intimately: the plot, character arcs, structural decisions and evolution of the prose. Ask why they made choices, how earlier drafts differed or why a character changed halfway through the writing process.

Passing off an AI-generated manuscript as original work is difficult under such examination.

Clear rules should outline whether a competition prohibits AI entirely or allows it for editing. Announcing an interview requirement would itself discourage people from flooding competitions with prompt-generated manuscripts.

Rather than walking away from the judging table, experienced writers and critics such as Chu are needed more than ever to uphold artistic standards.

The question of how to respond to AI extends well beyond literary competitions.

Chen, like many educators, also worries that students will misuse AI “to the point that their intellectual and creative abilities stagnate.”

General solutions often involve restricting AI while returning to handwriting, reading aloud and other traditional methods.

The concern is legitimate. Students can produce an elegant essay on the causes of World War I without understanding the causes or a sophisticated analysis of Macbeth without reading the play. A finished product can no longer reliably demonstrate how much intellectual work occurred inside a student’s head.

Imagine a history examination beginning with a proposition: “The atomic bombing of Hiroshima was necessary to end World War II.”

Instead of writing an essay, the student gets five exchanges with an AI instructed to challenge whatever position they take. Each response the students give must be at least 200 words. The student might argue that the bombing ended the war quickly and avoided the casualties of sending ground troops to Japan.

The AI asks: How do you know Japan would not have surrendered without it?

Each answer determines the next question. A weak argument receives one kind of challenge; a sophisticated argument, a more sophisticated one.

Could a student earn high marks without knowing the history, framing arguments properly and writing clearly?

The teacher would grade the student’s prompts, writing and reasoning, as well as how well they know the subject, put forward arguments, respond to counterarguments and defend or revise their position when better evidence emerges.

That is what might be missing from much of the debate about AI and education. The ideal examination has arguably never been the written test. It is a conversation with a knowledgeable teacher.

A university professor can still do it with a handful of students around a seminar table. Mass education cannot provide that experience to every student. AI potentially can.

For the first time, every student could have something resembling an individual examiner, capable of listening to an answer, challenging it and adapting the next question accordingly.

When AI becomes part of the examination itself, it stops helping students avoid thinking and starts making it harder for them not to think.

John Cheng is a retired businessman from Hong Kong living in Taiwan.