On college campuses around America, old-fashioned blue books are making a comeback.
Around a century ago these standardized booklets for written exams were introduced by Butler University in Indianapolis. For many former students of a certain age — myself included — the sight of a blue book can generate anxiety and even nightmares. And many former instructors — myself again included — recall the tedium and strained eyesight of trying to decipher students’ handwriting. So it was an improvement when exam-taking shifted from paper to computer. Or so it seemed at the time.
Now, however, students are using AI to write essays and answer questions on take-home exams, as well as taking in-person exams on their computers. As a result, an AI-arms-race has developed between instructors and students. Concerned that students are not doing the work themselves but are simply copying and pasting AI output, instructors have begun using AI programs to detect students’ use of AI. Inevitably, there are now AI programs that students can use to outwit the instructors’ AI detection programs.
So, not surprisingly, many instructors are going back to handwritten in-class exams, generating a sudden boom in the demand for blue books. Ominously, even the return of in-person testing may not solve the problem of testing in the face of AI: Cheating using AI glasses is on the rise in Asia and will doubtless spread worldwide.
My concern here isn’t about testing; it’s about learning. The objective of testing is to further learning, and there is growing concern (as well as evidence) that students’ use of AI damages their capacity to learn. And what we really mean by learning is the ability to think. Students who rely on large language models to answer questions won’t learn how to think by reasoning through the evidence to form a conclusion. As a result, they will be unequipped to deal with situations in which AI either can’t provide an answer or provides misleading answers.
In short, there are good reasons to worry that what we’re calling artificial intelligence will adversely affect the development of our natural intelligence. Moreover, in the case of basic learning, those adverse effects may be virtually irremediable.
The rise of generative AI isn’t a complete departure from an ongoing process of outsourcing human judgment and understanding to external models. Rather, generative AI is just a further step in a process that began a generation ago with the launch of Google search and accelerated with the rise of smartphones. However, ChatGPT and Claude Code ratcheted that process up to a much more rapid pace.
Granted, each stage of this process has brought obvious short-term benefits to those using the new technologies. Yet these benefits have come at the cost of real, measurable long-term damage to human understanding and cognition. And AI, which is already creating a crisis in education, will almost surely make the damage much worse.
Beyond the paywall I will address the following:
1. A brief history of outsourced cognition
2. The sharp deterioration in learning with the advent of smartphones
3. The AI cognitive crisis
4. Will cognitive losses due to AI lead to a new form of inequality?
