We are at a strange stage in the adoption curve of innovative technologies, where two seemingly contradictory things are true at the same time. In other words, it has become clear that artificial intelligence will transform the world. And the direct impact of this technology is still small enough that it's possible to pretend it's not actually the case.
Nowhere is this more evident than on a college campus.
Most of the assignments traditionally used to assess and, more importantly, challenge students, can now be easily outsourced to ChatGPT. This also applies to essays, the most classic assignments completed by students in humanities and social science courses. While the best students can still outperform AI models, the combination of advances in technology and rapid grade inflation means that students satisfied with an A- or even a B+ can safely cheat and graduate even from top universities.
The same is true of the primary assessment methods in many science courses. In fact, the AI models that earned top scores in math and science Olympiads may be even better at answering questions in problem sets in biology, chemistry, physics, and computer science classes.
For the most part, professors have responded by ignoring the problem.
Some people are in complete denial. Many academics and writers believe that chatbots are far less capable of performing a wide range of academic tasks than they actually are, due to real flaws such as the tendency to hallucinate that chatbots still have. Even though a significant percentage of students are submitting AI-generated work, they proudly reassure each other that their courses are too demanding or too human for any machine to understand.
Some people are well aware of the problem but don't really know what to do about it. If you suspect that an assignment was completed by an AI, it will be very difficult to prove it without confronting the student, which will definitely create a very awkward atmosphere and may even lead to a formal complaint. And even if they can somehow prove a student's wrongdoing, there is a long and frustrating bureaucratic process that can end with university administrators imposing very light penalties or telling professors to turn a blind eye.
Another form of assessment may be a solution. However, oral questioning and face-to-face exams using pen and paper have become obsolete. These are likely to anger students and require a lot of effort to manage anyway. So even for those who are aware of the problem, the path of least resistance is to continue pretending it doesn't exist.
An old Soviet joke went: “We pretend to work, but they pretend to pay us.” In many universities today, students simply pretend to be doing their studies. For now, most professors are still avidly reading and commenting on ChatGPT efforts. But I think some of them will decide to outsource grading to artificial intelligence as well. When that happens, campuses will reach a new stage of AI decadence. Students pretend to do assignments, and professors pretend to grade assignments.
I won't deny it, However, it is forever an option.
Technology will continue to evolve in the coming years. Students who have used AI tools throughout their high school careers will begin flocking to campus. They will become much more proficient at completing traditional challenges using these tools. You might even become adept at using these new tools to accomplish some truly impressive things. The argument that current assignment formats have any meaning or that college GPAs give employers meaningful signals about candidate quality will become untenable. At the same time, some of the fundamental skills that students need to acquire to truly understand their chosen field, or simply to become fully formed citizens who can reason carefully about the world, will rapidly atrophy.
What should universities do about this? Is the right path to fully embrace AI tools, or to take more drastic precautions against their widespread use?
I've come to think more and more that the answer is both.
whoever wants it To make a real contribution in the future, whether in the workplace or in academic research, we will need to make good use of AI tools. Therefore, it is the job of universities to teach students how to make the most creative use of these tools, which many students are currently unable to do.
But even in a world where AI tools become increasingly powerful and pervasive, fundamental skills like clear thinking and powerful writing remain critical. This means that AI tools make it easier for students to bypass the difficult work required to master these skills, which poses a real threat to intellectual growth.
Writing an essay can feel like a very artificial exercise. And of course, we're entering a world where many of the writing tasks once associated with white-collar jobs, from emails to business plans to PowerPoint presentations, can be outsourced to AI as easily as college assignments. One might be tempted to conclude that academic skills that were once very important, such as the ability to write, have lost their importance.
But this ignores the point I've been stressing to my students long before capable AI models were released: Writing is thinking. When giving a talk, it's easy to be vague about ideas you don't fully understand, or to skip some logical steps. The moment you try to commit your argument to paper, its weaknesses are mercilessly exposed. (Actually, that's why I can't believe that people who claim they're just bad writers are being honest with themselves. In most cases, bad writers are bad at writing because they don't make the effort to think through their ideas.)
If you want to be a successful artist today, you probably spend little time etching still lifes or creating works that involve difficult problems with perspective. However, most art schools still recognize that mastering these skills is a necessary part of education. The same goes for skills like writing, which could theoretically be outsourced to ChatGPT. Although you may not need to rely on those skills directly after graduation, mastering them will give you skills and habits that will enable you to better understand and act in the world.
This requires universities to place greater emphasis on both foundational skills and the use of new technologies than they currently do. The students who can contribute the most in the future are those who are forced to write a large number of traditional essays without using digital tools. and Someone skilled at using AI to push the limits of human knowledge
at the moment, Universities are choosing a dangerous middle ground. Instead of leaning into the possibilities that come with leveraging the incredible capabilities of AI, we cling to old forms of evaluation as if they still have any meaning. Instead, we need to dichotomize different forms of assessment. In some courses and situations, students must be forced to prove their intellectual mettle without the use of digital tools. Other courses and situations should provide the knowledge and know-how to use these tools to maximum effect.
This is something I hope to experiment with myself when I teach two undergraduate seminars at Johns Hopkins University next semester. For the first time since I started teaching at the university, we will be conducting in-person exams that require students to answer with paper and pen. They have three hours to write three essays on the broad themes of the course, demonstrating mastery of the material and ability to make persuasive arguments without external assistance. But for the final research paper, the cornerstone of a rigorous undergraduate seminar, I encourage them to use AI liberally. They need to know and document exactly how they use AI to support that project, but I evaluate them solely based on whether the final product makes a meaningful intellectual contribution.
The most skilled pilots can fly a simple Cessna with very little technology involved, or they can handle the myriad of equipment contained in a Boeing 787. Similarly, the most well-prepared workers, scholars and future citizens will be able to think for themselves without the help of ChatGPT and, if necessary, will be able to skillfully seek the help of such a magician's apprentice. Our job as their teachers is to help them achieve both.
This essay is short contribution Visit the Chronicle of Higher Education's forum on AI and education.
