The Ghost in the Schoolroom: How Early AI First Taught Our Children

When we discuss ChatGPT and the integration of Large Language Models (LLMs) into classrooms today, we tend to believe we are witnessing an unprecedented technological revolution. However, the history of Artificial Intelligence (AI) in education has deep roots, stretching back more than half a century to university labs and experimental schools where visionary scientists sought to encode the process of learning itself. Seeking the answer to where this early form of AI taught children leads us to an era where computers occupied entire rooms, yet the ideas for personalized instruction were already remarkably sophisticated.

Seymour Papert’s Turtle and the LOGO Language

In the late 1960s at MIT, Seymour Papert, a mathematician and psychologist who had studied under Jean Piaget, began to challenge the way computers were being used in education. Rather than the computer "programming" the child, Papert believed the child should program the computer. Thus, LOGO was born—the first programming language designed specifically for children. Its most recognizable feature was the "Turtle"—initially a physical robot on the floor and later a graphic on the screen—which children guided using logical commands.

This approach, known as constructivism, was not merely a coding exercise. It was an early form of "embodied" AI, where the child learned geometry and problem-solving by teaching a machine how to move. These experiments took place in public schools in Brooklyn and Dallas, proving that technology could unlock creativity rather than restricting it to automated testing. Papert’s vision was about creating "microworlds" where children could explore complex ideas through play and logic.

The PLATO System: The Forgotten Giant from Illinois

While MIT focused on individual creativity, the University of Illinois was developing PLATO (Programmed Logic for Automatic Teaching Operations). PLATO was, in many ways, the ancestor of the modern internet and educational AI. As early as the 1970s, the system featured touchscreens, plasma displays, and a network that allowed thousands of students to share lessons simultaneously. AI in PLATO manifested through "branching logic," where the system evaluated a student's answers and dynamically adjusted the difficulty of the next step.

PLATO did not just teach mathematics and languages; it created the first digital community. Students and teachers used the system to send messages (early emails) and participate in discussion forums (notes). It was a testament to the fact that AI in education is most effective when it serves as a bridge between humans, rather than just a digital substitute for a tutor. The system’s longevity—running for decades—showed that public-sector innovation could set the standard for the private market.

Intelligent Tutoring Systems (ITS) and the Socratic Model

In the 1970s, research shifted toward Intelligent Tutoring Systems (ITS). The SCHOLAR system, developed by Jaime Carbonell, represented a major breakthrough. Instead of following a pre-written script, SCHOLAR used a semantic network of knowledge to conduct a dialogue with the student. If a student asked a question about South American geography, the system could "reason" and respond based on the relationships between data points, mimicking the Socratic method.

These systems were tested in military academies and university departments, laying the groundwork for what we now call adaptive learning. The challenge then, as it is now, was "student modeling": the machine's ability to understand not just what a student knows, but why they are making a mistake. This psychological dimension of AI remains the "holy grail" of educational technology, requiring the machine to possess a theory of mind regarding the learner's misconceptions.

The Political and Social Context of Early AI

The development of these systems did not occur in a political vacuum. Following the launch of Sputnik, the U.S. government funneled massive amounts of funding into educational technology through DARPA and the National Science Foundation. There was a profound hope that AI could bridge the inequality gap, providing every child with a "personal tutor" that never tires. This was the "Great Equalizer" narrative that still dominates EdTech discourse today.

However, history teaches us that technology alone is insufficient. Many of these early systems failed to achieve widespread adoption due to the prohibitive cost of hardware and resistance from traditional educational structures. Today, as GenAI enters schools, we must look back at these experiments. Early AI taught children not just how to solve equations, but how to interact with an intelligence of our own making. The lesson remains: AI in education must be a tool for empowering the human spirit, not a mechanism for automated compliance.