Will AI transform education? The case for
Previous posts looked at AI in education and then at the case against the view that AI will transform education. This post explores why AI might in fact revolutionize education.
The New York Times published an expansive essay, “What Would Plato Say About ChatGPT?”, that includes this provocative opening:
“Plato mourned the invention of the alphabet, worried that the use of text would threaten traditional memory-based arts of rhetoric. In his “Dialogues,” arguing through the voice of Thamus, the Egyptian king of the gods, Plato claimed the use of this more modern technology would create “forgetfulness in the learners’ souls, because they will not use their memories,” that it would impart “not truth but only the semblance of truth” and that those who adopt it would “appear to be omniscient and will generally know nothing,” with “the show of wisdom without the reality.”
If Plato were alive today, would he say similar things about ChatGPT?”
I believe that quote holds the key to why AI might be the technology that actually does transform education. Explaining why will take a few paragraphs.
The affirmative argument is based on a few key points:
First, let’s acknowledge that the previous main view about AI in education, which is that AI would transform learning via personalization as embodied by Knewton and other companies, has been mostly a failure to this point. That’s not the main channel of the transformation path at this time, although it could be an important factor in the future.
Second, recognize that some of the shortcomings of the currently public available versions of ChatGPT have probably already been fixed in the current private versions. Arguing about fixable shortcomings of the current systems is a waste of time when we’re considering future impacts. (These shortcomings are important to understand relative to current usage, however.)
Third, let’s think about what these AI systems might be able to do in the relatively near future, combining ChatGPT, Dall E, and some other systems and capabilities.
Much of the current focus is on Chat GPT’s ability to generate essays. But when you recognize that capability as largely a synthesis and summarizing capability of a significant percentage of all the information known to humanity, and combine it with both natural language processing and image generation, what do you have?
A system that could drive the marginal cost of online content generation down to essentially zero.
What about the need to align to state standards and demonstrate alignment? Again, the cost would be roughly zero once the system is created.
Why is this so transformative? Because it could change the fundamental interactions between content/instructional materials, teaching, and other elements of student support provided by school-based professionals (SEL support, guidance, career and college, etc.)
There’s been a concept in online learning for a very long time, that good online content and platforms allow teachers to focus on tutoring, relationships, and SEL-related issues. That’s true to some extent, but as long as content generation has been relatively expensive, and the personalization potential of online platforms not yet fulfilled, there’s not been true cost savings to allow significant investment in these roles for teachers and other professionals.
Imagine a world in which AI systems have been trained on the best open educational resources (with appropriate copyright considerations, etc) to create:
Standards-aligned content that is consistently higher quality than is currently available (because there is excellent digital content available, but it’s hard to source from a single provider and/or it is expensive).
This content is fully aligned to all state and other academic standards.
It includes text, videos, assessments, animations, etc.
It sits in a system that can assess student understanding and personalize content.
Its marginal cost is close to $0. And by marginal cost I don’t mean marginal cost to deliver to one more student, but marginal cost to create a new course.
The first four bullet points don’t sound too different from what we are often told is currently available, but AI could take the production values, accuracy, and overall quality to an entirely new level, while finally delivering on the personalized learning aspects that previous AI systems promised but couldn’t pull of in practice and at scale. But it’s the drastic reduction in cost that’s really the key.
This futuristic system would make teachers obsolete, right?
Not at all.
This is where we get back to the above New York Times quote about Plato.
In an AI-driven education system, teachers’ roles would shift away from delivering content, to working with students instead. Their work would focus on understanding students, motivating them, addressing SEL needs—and deeply assessing their knowledge, understanding, and interests.
We’re already seeing some of this shift in online and hybrid schools, in which teachers are already filling a different role than in traditional schools.
But the AI-driven version of content and personalization could dramatically accelerate this shift.
There would also be the need for a significant increase in teachers assessing student knowledge in a world in which any student can create an essay, slide deck, etc. in minutes using AI.
What does that assessment look like?
It will likely include much more use of strategies that look like a version of the Socratic method, or a dissertation defense: one or more teachers discussing key issues with a single student, or a small group of students. Of course, this can’t happen just at the end of each school year; it needs to happen often and consistently. Assessment and instruction would become far more intertwined than they are in most traditional classrooms.
To be clear, I don’t know where this would lead. But it would take schools down a path that has been talked about by digital learning advocates, and we see to some extent in online and hybrid schools, but has not yet reached scale—a path in which the roles of teachers and other professional staff become very different.
How and why would it reach scale?
I doubt this approach is going to take hold in traditional public schools in less than a decade even in the most optimistic scenario—and it might far longer if traditional public schools were the only path. Too many elements of public schools are too difficult to change quickly.
But taking a page from the Christensen Institute’s disruptive innovation theory, you can envision how a set of charter schools, alt ed, and independent study could be built around new instructional models taking advantage of the changed cost structures, and newly enabled roles, driven by AI.
Again, this isn’t so different than the promise of online schools. But there are two major differences. First, fully online schools have never been of interest to a significant percentage of students and families, for all sorts of reasons linked to (real or perceived) socialization, and the day care role of schools. Second, existing cost structures have made innovative hybrid schools difficult to scale. This time could be different due to the changed cost drivers explained above.
A friend of mine with long experience in Silicon Valley talks about how a key question about AI might be: is the timing similar to the Internet circa 1989, or 1999?
If you’re old enough to remember those times, you may recall (or have realized later) that the Internet existed in 1989, but it was too early to have much impact. It was hard to use, and usage was limited to a few intrepid virtual explorers testing the earliest listserves, chatrooms, and eventually, browsers.
By 1999 the Internet was exploding all around us, on a near-term path to becoming ubiquitous and changing most elements of life. A hype bubble formed, but even the bursting of the bubble contained the seeds of the next transformations such as smartphones and social media.
Is AI the technology that enables the changes that transform education? Is 2023 more like 1999 than 1989?
I don’t know. But after 25 years about being skeptical of new technologies, I’m actively considering that this time might be different.