Editor’s note: This essay is an entry in Fordham’s 2023 Wonkathon, which asked contributors to answer this question: “How can we harness the power but mitigate the risks of artificial intelligence in our schools?” Learn more.
One of education’s most iconic design breakthroughs is a simple wooden shelf.
Students can move freely about the Montessori classroom to select work that interests them. The materials on the shelf are carefully crafted to meet pedagogical goals. Puzzle piece pegs are sized to train young hands in the fine motor skills they need to hold a pencil. Tactile letters help students build phonemic awareness and learn to write without writing.
Students can move about the classroom, free to make their own choices. Teachers will provide instruction and support, but some of the most important instructional choices have already been made in their construction of the “prepared environment” and their curation of the work on the shelf.
The puzzle that has vexed generations of Montessori educators is how to extend this same design philosophy to the education of older students who have outgrown the nursery.
Maria Montessori laid out a vague, high-level vision for secondary education in a series of appendices to From Childhood to Adolescence. The appendices describe students living in the countryside, reading about the history of humankind, tinkering with “machines,” and helping run actual businesses, like a bed and breakfast or a working farm. They describe itinerant subject-matter experts who would stop by these idyllic settings and deliver seminars.
All over the world, learning environments attempt to translate these high-level principles into twenty-first-century practice. Students need not decamp to the countryside. They can work at a nearby café or bike repair shop, as they do at Embark Education in Denver. The tinkering projects, historical inquiry, and itinerant experts can all be curated in digital learning software and brought to life through in-person learning experiences inside the classroom or beyond the school walls.
Ray Girn, the CEO of Higher Ground Education, has described how his company is setting out to help educators assemble “the digital equivalent to a Montessori shelf, with all the pedagogical innovation and richness such an approach implies.”
Just as a primary student can freely explore a Montessori classroom and select from carefully crafted activities, an older student can access an online platform, where they can browse not just a digital playlist of projects and assignments, but a host of internships, apprenticeships, community service projects, local theater productions, summer camps, college courses, tutoring providers, and countless other learning experiences that might exist outside the four walls of school.
ReSchool Colorado, with its Blueprint4Summer Initiative, assembled details about thousands of summer learning experiences into a database families could easily search. What would it look like to scale this effort to every community, and extend it to the full range of learning opportunities that might interest students throughout the year?
Taming the data about all these learning experiences, and helping educators assemble and curate it in a way that allows them to present it to students in an experience as seamless and elegant as a Montessori shelf, could be a job for AI.
In practice, most recent discussions of “artificial intelligence” refer to software that is capable of interpreting commands delivered in plain human language and generating novel output.
Large language models like Open AI’s GPT-4 or Facebook’s Llama have obvious applications in education. They may allow software to grade not just multiple-choice questions, math problems, or other assignments with pat right-or-wrong answers, but more ambiguous work, like student essays. In addition to evaluating whether an answer is right or wrong, they could help diagnose errors and deliver instantaneous feedback.
A less-discussed application involves assisting human educators to interpret ambiguous and not-at-all-standardized datasets, like the range of learning activities an adolescent might choose for themselves.
A growing number of organizations are designing learning environments to connect students with learning experiences outside their walls. Students in the Cristo Rey Network’s work-study program spend one day a week working in local offices. GEO Academies sends students away from campus to enroll at local colleges or career training providers. In Idaho, One Stone helps incubate student-created companies. Online schools like Tech Trep Academies serve as brokers of learning experiences provided by others.
Education savings accounts have the potential to make similar opportunities available to all students, with or without the school serving as an intermediary. Students can use their accounts to pay for a few online courses, transportation to an internship, and an in-person tutor. Florida has created a new support system for these families, known as “choice navigators.” These individuals help students set learning goals and select the range of learning opportunities that will help them meet those goals.
One unmet challenge is equipping these navigators with actionable data on what all those potential opportunities look like. It’s impossible for one person to hold all this information in their head, and it’s difficult to imagine a neat and standardized database assembling everything from one-off projects and online curricula to internships and tutoring centers in a coherent way.
Making sense of diverse learning opportunities to give actionable advice to each student requires the individual touch of a human and the raw data processing power of a machine. AI can facilitate translations between the two.
This would be a critical step toward developing what Northwestern University’s Nichole Pinkard calls “connective tissue”: the informational and physical infrastructure that helps families and educators connect all students to all the different experiences, in and out of school, that help them meet their learning and developmental goals.
I’m talking here about deploying AI to help break down data about education providers and learning experiences. I am not talking about student data. That is an area where technologists looking to build these tools must proceed with caution.
At a minimum, people working in education should listen carefully to the more-advanced debates in the medical field around the ethical implications of AI and the need to protect patients’ privacy.
Even those of us who are generally skeptical of the need for regulation of AI and worried that the Biden administration’s efforts to “guide” the development of these technologies will ultimately limit their potential should agree that protecting the privacy of minors is one area where thoughtful regulation is likely necessary.
The goal should be to allow every young person to freely, safely, and easily navigate a range of diverse learning opportunities that are crafted with the same elegance and simplicity, and curated by human educators with the same pedagogical deftness as the activities on a Montessori shelf.
