National Test Scores Are Down, Is Generative AI Partly To Blame?

American middle and high school students are not performing well on assessments, indicating that fewer of them are learning even as much as their peers were just a handful of years ago.

Being fair, the educational attainment bar of a handful of years ago was not especially high. Based on our national test scores, educational mastery, competency, and preparation of middle and high school students has been slowly and inexorably crumbling for some time. Even so, our most recent national test scores – the nation’s education report card – are “at historic lows.”

In the past, I’ve written several times about how we should not panic about big assessment scores. But I am starting to worry. No, given this new data, I am worried.

For just one example, barely more than one in three high school seniors (35%) is prepared for college work. That’s down from the meager 37% it was in 2019. Another – just 22% of American seniors are “proficient” in math. Not exceptional or even average. The marker is proficiency and not even one in four students is meeting it.

Knowing this, seeing the state of decay in our national and international scores, reasonable people can ask only two questions – what is causing this, and what can we do about it?

The last part is easy. Everyone knows how to improve educational performance. It’s no great mystery. We just don’t care enough to actually do it.

As for what’s causing this decline, what’s behind this most recent downward acceleration, most people have marked an obvious culprit – Covid. Fair. And no doubt true.

But curiously, no one in this responsibility conversation is even talking about the other thunderclap to shake our schools in recent years – generative AI. It’s a baffling omission.

The tests for which we just received scores were given between January and March 2024. ChatGPT came out in November, 2022. And we know that since then, students have been uniformly using it to outsource their school work, to offload the work of learning.

Even putting the cheating shortcuts aside, everyone seems to agree that OpenAI/ChatGPT and its cousins have “changed everything.” For the past few years, millions of people have been gaga breathless over how Gen-AI will require rethinking, redesigning, reimagining, reengineering – pick your favorite re-word – everything about education.

Yet, when our national test scores go down, this transformative new technology is absent, just excused from the conversation. Frankly, I don’t think it plausible that something as supposedly game-changing and ubiquitous as generative AI has played no role whatsoever in these recent test scores. It can’t be everywhere and everything, and the moment something goes wrong, also be nowhere and nothing.

Is it not at least possible that the double disruption of Covid and ChatGPT played some role in our most recent assessment embarrassment? Does generative AI have any hand in this debacle?

Honestly, I’m not sure. And I’m not sure how we’d know, even if people were asking, which they do not seem to be. So, I put the question to an open forum of education experts – did the explosion of ChatGPT and other generative AI tools play any role in the most recent test score declines?

Most of the experts said no. They seemed pretty confident that generative AI was not any significant factor in our dropping performance scores. Moreover, most of the people who said no also said AI was some version of apple pie amazing – that schools had to stop fighting it, and embrace it, and so on. Most said AI was a force of good and did not belong in a question about bad things.

But not everyone felt that way.

Joe Vercellino was the 2021 Detroit Teacher of The Year and is the author of “Toiling With Creativity,” a book about teaching, teens, and AI. He was not shy, saying, “Artificial intelligence will create a stark contrast between those that intentionally develop skills to understand and those that pawn off the need to build competence to simply solve.

“The dependence on AI to solve problems that young people need to understand will constantly undermine test scores and performance in real-world critical thinking scenarios,” Vercellino said.

Karen Aronian saw the AI/test score link too. She’s a former NYC public school teacher, and college professor at private and public universities across the United States, with an education doctorate from Teachers College, Columbia University. “AI is replacing students’ investment in their work and critical thinking. Especially for the seniors who have one foot out the door of high school and juggle so much, AI is a quick fix that they can easily become addicted to,” she said.

Looking at the window between March 2020, when Covid functionally ended actual education and instruction, and March 2024, when these recent tests ended, middle and high school students have had maybe a single year of normal school without AI everywhere. In those four years, students have not had to do much schoolwork – at least not if they didn’t want to.

I’m not making a case the generative AI sank our national test scores, the reflection of educational success. Again, most of the experts I asked did not think that was true. But it ought to be something we consider, that it’s at least possible that one thing is causally linked to the other. And if we allow for that possibility, it just might inform how our education leaders in classrooms, in district offices, and on school boards consider and implement policies about AI use in education.