Overview:
We're rushing into AI in education without seriously examining the risks, and given past cautionary parallels (social media, digital classrooms), we should slow down before real generational harm is done.
I see a lot of people on LinkedIn (and other places) writing about how potentially great A.I. is, personalized learning, making learning fun, allowing expression, discovery, and a host of other potential benefits. Some benefits are being realized. And I agree with many posters in principle, as a paying ChatGPT-certified user, a writing-with-A.I. teacher, and a parent. One major thing missing in the hype is balance.
What if we get A.I. in education wrong? Why should we be moving slower? When would we know the problems? How would the faults appear? Who will be affected? What else is there? Glimpses are appearing now, with homogeneous writing and even design elements.
Are these even the right questions? When A.I. in education has gone wrong, it is too late, and some damage has been done to the education and development of young minds. A generation of clones. Prevention is better than a cure. Or should we be asking, “How do we prevent A.I. from going wrong in education?”
There is a lot written about Alpha School and others that use A.I. to engage and support students. But these are small-scale examples with involved parents (or at least parents who can pay for support). Scaling is rarely discussed. I have asked on several LinkedIn posts, without much response. Charter schools held promise, too, but without support, many graduates who shone in high school dimmed greatly at university (when controlled for factors like selection, home, etc.). A mixed bag, similar to what is happening with A.I. schools popping up now.
Examples are given of making homework fun and how promoters are using AI with their kids. True, A.I. can make homework fun, engaging, and learning. However, not many parents are actively involved in their child’s education. Of the 16 kids in my son’s 4th-grade class, only two of us are actively involved in our kids’ learning. Even then, my son says that I am doing too much and just writing a simple prompt is okay because that is what his friends are doing! Among my university classes, less than 1 in 5 students report parental involvement in their education. As the age-old saying goes, friends influence more than parents.
We have seen with digital classrooms that learning can be accessible but not optimized, or even limit the internalization of knowledge. There is an appearance of knowing without truly knowing. A 2024 study showed that students who use A.I. assistants in studying do worse on tests. No wonder Sweden limited its digital classroom and brought back more book time.
Social media, too, had significant promises, but has resulted in significant mental and even physical problems among youths. Australia’s band and other countries’ limitations might seem extreme to those of us who grew up pre-2000, but those generations (X, Y, and even some boomers) know how to handle technology much better. They experience its evolution. A teenager in 2025 has had the internet and social media for their entire life. The only change is the platforms. A.I. appears new, but it is a continuation (or part) of machine learning.
Hence, some of the mature proponents can use A.I. to achieve enormous benefits. Students with disabilities or requiring language assistance can also benefit greatly from AI. But there are many being left behind. Those who use LLMs for direct answers rather than as a thinking tool, or even thinking about the answers. It is not the technology per se, however, reliance to dependence is a short line. Usage is crucial.
Success in education is measured in generations at the societal and individual levels. We are seeing the soft parenting in the 1990’s reflected in the extreme gentle parenting of the 2020s. Kids as accessories. Combined with the pervasiveness of technology that can act as babysitters (from TVs with many channels to tablets with YouTube), parents choose not to spend their time reading to kids. Declining reading rates have accelerated in the past 10 years because those 1990s kids are now becoming parents with more toys to direct attention and responsibilities.
The rise of block learning in the 1980s – 1990s has led to major policy problems now. Semesters became common, and courses were revised into varied chunks. Gaps were ignored. As a result, seemingly intelligent people (tertiary educated) falter on simple concepts. For example, climate is long-term (usually 30 plus years), and weather is short-term (usually less than a year). This high school concept has been missed by many. So, when a PhD expert says climate change is a hoax, it misleads those with less qualifications. The so-called expert lacks a comprehensive understanding but can influence policies that are not good for the same masses. The reign of confusion, even among intellectuals, started many years ago, and has gotten worse. And might get worse as preliminary studies show AI is rewiring our brains!
Those who challenged the changes, as usual, are called antiquated. Holding back progress to maintain their fiefdoms. There is, like all things, some truth in these statements. In all things, there are good and bad. Balance is what is needed. The Luddites did not resist technology, but the implementation.
However, after more than 150 years of public education, should we, educators, not reflect on the long term and use the great tools we have to perform a balanced assessment? A recent OECD report said that Edtech has not delivered on its promises (especially to the masses) after 10 years of widespread implementation. Basic literacy and numeracy skills are needed to master AI tools. What’s that saying about putting the cart before the horse?
While I am using A.I., teaching my students to use A.I., and involving my kids with A.I., I am also a pragmatist. Parents, schools, and society need to ask questions about what if we get A.I. in education wrong. There are many recent examples of getting things wrong. There is no need to recall eugenic and mental health studies conducted 100 years ago (though they are enlightening reads). Smoking was once promoted as healthy and ignored for generations. Walking barefoot in the grass with animals was once considered backward and unhealthy, but is now known to be physically and mentally beneficial. A litany of misplaced hope examples exists.
So, with just over three and a half years of ChatGPT and widespread use of A.I., we must take some time to think and give direct answers about the direct and generational impact of getting A.I. in education wrong.
This is right on the money. I know a lot of people earning big bucks from edtech companies to fly around and present about how great AI in classrooms is. I heard the same thing about cell phones. It didn’t work out.