Getting the Most From AI

Ever since that fateful day in November 2022 when OpenAI released ChatGPT, hardly a week has gone by in which AI hasn’t been in the news. Much of this coverage concerns fears of AI-perpetuated labor-market disruption and mass unemployment. A look at the theory underpinning these fears can help us understand what may happen as AI’s reach increases. More importantly, it can help us understand AI’s threat to education and what colleges can do to turn this threat into an opportunity.

Capital—a term referring to assets used in production—can either complement or substitute for labor. As a substitute, it replaces workers by doing what they used to do. As a complement, it makes workers better at what they do, allowing a company to do the same work with fewer people. Interestingly, capital can simultaneously be a complement and a substitute. ATMs replaced much of a teller’s work, making his or her remaining tasks more valuable and lowering the costs of running bank branches. As a result, more bank branches were built, and teller positions skyrocketed.

AI, like the ATM, is a potential complement and substitute for labor. It is the same for learning. AI, like the ATM, is a potential complement and substitute for labor. It is the same for learning. AI can complement someone’s learning by granting him or her access to highly effective tutoring and by making his or her existing knowledge more valuable. Or it can substitute for education, allowing students to avoid learning by having AI do everything from writing to reading for them.

Colleges must teach their students how to use AI without letting them forsake learning. In Democracy in America, Tocqueville called attention to one of the dangers of soft despotism: the diminishment of capabilities that people experience after outsourcing certain responsibilities to the state. AI is another potential source of soft despotism, as, if misused, it destroys our ability to do things such as write, think, and code over the long term. This is especially true for those still learning, as, while others may stop doing things they’ve done before, current learners may never learn how to do new things.

In threatening learning, AI also endangers colleges’ ability to educate students and prepare them to contribute to a free, flourishing society. Kids who never learn to do basic things independently are not just uneducated, they are poorly suited to participate in the workforce and society. This is an existential threat to every college and one they cannot wave away by avoiding a choice. No matter what colleges do, students will naturally succumb to the desire to use AI. Moreover, AI skills will soon be critical for many jobs. Thus, colleges that fail to teach AI do their students a real disservice.

In the face of these challenges, colleges must teach their students how to use AI without letting them forsake learning. The way to do this is by adopting human-autonomy-centric approaches to AI education—similar to those that math educators have adopted in the age of calculators. Such an approach would start by equipping students with foundational knowledge and teaching them how to apply it. After their competency is guaranteed, students will be shown how to use AI and how their education can be used to get the most out of it.

For example, a computer-science student can be taught the core concepts of programming basic apps and functions. Once students have had an opportunity to build digital applications on their own, educators can introduce AI and show students how to use it. By demonstrating how knowledge of computer science lets students get better code from the AI they are using and adapt that code for their specific situations, these educators can help students use AI to complement their knowledge rather than replace it.

By adopting this approach, colleges can prepare their students for the workforce without compromising other parts of their mission. By showing the continued value of learning, they will even spark students’ interest in learning more by showing them the continued relevance of education.

AI is poised to reshape many parts of the world in different ways. Education is at a crossroads, capable of being changed for the better or worse. By embracing autonomy-centric approaches to AI, universities can ensure that AI improves education by helping students get more out of their knowledge rather than keeping them from learning in the first place.

Zev van Zanten is an economics and mathematics student at Duke University.