AI and the Incentives We Never Updated

Kang’s recent column in The New Yorker frames a question many parents are asking in private: if higher education is primarily about credentialing students, what prevents students from using AI to gain a leg up in credentialing their resumes at the expense of their own lesser thinking?

Kang summarizes that higher education is still primarily used as a signal for higher thinking, work ethic, and employability. Given this incentive signal, the old reward system remains in place: admissions reward polished papers, CVs stacked with activities, tests and academic metrics that read well on paper. Scholarships, internships, and first jobs still lean on those archives.

And it sort of worked without AI. Achievement took real substantial effort and time to master. But what did change in the last several years is the cost of producing the artifacts that ladder rewards.

High school students face the same pressure to look exceptional and now they have tools that can draft, revise, tutor, and simulate expertise at a pace original thinking cannot. The rational response under the old incentives is not “think harder in solitude”, it will often shift to “use AI” to clear the bar faster and with less friction.

But the tradeoff is undeniably terrible for the AI student (and also their thinking counterpart) in the long run.

Reciting an AI-generated answer is not the same as wrestling an idea for hours until it’s polished. Getting an essay to “sound elite” is not the same as learning to structure an argument when you are confused.

On paper, the shortcutters will be rewarded. They look like the strongest applicants with quick logical clear thinking and reasoning. The students who toil in their raw thoughts, fail frequently, and take longer to reach clarity are ultimately punished. Ironically, the independent but more developed thinkers actually gain less access to the higher level thinking they deserve.

These incentives have to change if we prize individual and original thought. How will society learn to decipher true thinking from AI thinking? It’s impossible for students and teachers to avoid using AI, but we need to think about how credential stacking used as incentives will ultimately limit what truly makes us original.