Everyone Hates AI. Where Did Schools Go Wrong?

AI frustration
(Image credit: Getty Images)

It’s hard to overstate the feeling of excitement around artificial intelligence in education in the early ChatGPT era. Sal Khan told me in February 2023 that AI was not only a new epoch but “the most epoch of the epochs.” Other educators were equally enchanted by its promise to save teachers time and provide on-demand personalized feedback to students through AI tutors.

A few years later the feeling around AI use in schools has shifted dramatically. Critics of the technology have grown more vocal and more numerous. A family in North Carolina is suing Google, alleging that their 13-year-old daughter had conversations with AI on a school-issued Chrome device that contributed to a later mental health crisis she endured. Authors are losing literary awards and publishing deals over alleged AI use, and everyone is tired of all the AI slop around, including in published academic papers. Outside of the classroom, heightened fear over AI safety and other concerns about the technology have made AI datacenters extremely unpopular.

To be clear, there are still strong AI proponents in education who can correctly point to many beneficial uses of the technology including vibe coding, research, and accessibility features such as translations and speech-to-text features.

Still, even staunch proponents have had to wrestle with limitations of the technology. One of the main challenges with Khan Academy’s AI tutor, Khanmigo, for instance, is that students don’t seem motivated to use it. In addition, AI proponents now have to navigate parents and a public increasingly wary of AI, complicating future implementations of the technology.

It didn’t have to be this way. While some schools took thoughtful and cautious approaches with AI, overall, the education field blundered with its implementation of AI. Many foreseeable mistakes could have been avoided.

Here are what I see as the biggest mistakes made. The good news is there’s still time to correct course and use AI better with less of a downside.

1. Not Taking AI Cheating Seriously Enough

This is the original sin of AI use in schools. From the get-go, the potential impact of generative AI on cheating was clear, and not enough was done at the institutional level to prevent this. Instead, schools seemed to surrender to AI slop without a fight, essentially conceding that from now on human-generated academic writing was a thing of the past.

What makes this particularly frustrating is that there are ways in which schools could have, and can still, fight back against AI cheating. Plagiarism detectors aren’t perfect, but some get pretty good results. With the right student protections in place, these could be an option.

AI watermarks are starting to become more common, and educational leaders should have been calling for these sooner. Now they should make sure they have clear plans on how to use this feature going forward.

Humanity in writing and student work is important, and that authentic human voice also should have been centered in the conversation around AI writing from the start.

2. Failure To Measure Efficacy Well

I was impressed with some AI tutors early on, but there’s a difference between a tool that seems to work well for an adult and one that students will actually use.

Motivation is a huge part of learning, and I haven’t seen many generative AI tools that are particularly helpful in this regard. Too many institutions seemed to take it on faith that AI optimization equaled real optimization.

We need to be honest about AI tools and their abilities, and acknowledge that AI still makes a lot of mistakes and therefore might not be ready for prime time.

Mistakes aside, schools should not be investing large amounts of money into technology their students have little or no interest in using, as many students have developed ethical objections to generative AI.

That said, I agree that AI tutors and other AI technology has tremendous potential. That’s why I want AI tools studied and tested rigorously. We have to ensure we’re getting the most out of it for our students.

3. Not Learning Lessons From Phones

Cell phones in the classroom were embraced by some educators over the past decade or so, but there’s increasing consensus that these don’t belong in the school setting.

I have experienced this myself. I used to allow students to have their phones in class, figuring that it would help them be active and look up things. Instead, phones ended up being a major distraction, and I had to change my policy and start being much stricter about their uses in my classes. Many school districts experienced a similar evolution on a larger scale.

Now, the pattern is being repeated with AI. Like smartphones, there is clear educational potential to AI, but there is also the potential for distraction, and worse, harm to students’ mental health.

The relevant lesson from school cell phone use is this: just because a technology is widespread and likely inevitable, it doesn’t mean it automatically belongs in the classroom where it can be harmful to students.

We could have taken a more cautious approach with AI than we did with phones and made sure to carefully and thoughtfully vet its presence in the classroom. Instead, too many schools jumped on board the AI train too soon.

4. Not Enough Focus On What AI Actually Does Well

All the negative aspects of AI use have unfortunately overshadowed some truly great use cases in education and beyond.

AI translations and speech-to-text transcriptions have done wonders for accessibility in education, and have the potential to help much more. Additionally, AI can be a really helpful research partner and one students should be learning to harness. I’ve not personally found AI assistants or agents to be that helpful for my work but I know people who love them, and there are many more positive AI uses.

These are the types of AI wins educators should be building on, however, to do that effectively, we need to be more critical about what is and isn’t working and about the downsides of AI technology. In doing this, as I said before, we need to always make sure to center student welfare and humanity in the classroom.

AI is a tool, and a potentially helpful and even transformative one, but it should never replace human thinking, writing, or creativity of any kind. Once we’ve firmly protected humanity in our schools and classrooms, we can start finding more pathways toward effective human and robot collaboration.

Erik Ofgang

Erik Ofgang is a Tech & Learning contributor. A journalist, author and educator, his work has appeared in The New York Times, the Washington Post, the Smithsonian, The Atlantic, and Associated Press. He currently teaches at Western Connecticut State University’s MFA program. While a staff writer at Connecticut Magazine he won a Society of Professional Journalism Award for his education reporting. He is interested in how humans learn and how technology can make that more effective.