Slaves to the Machine

In 1966, a Stanford University professor promised to harness the power of the computer to provide “the personal services of a tutor as well informed and as responsive as Aristotle.” In 2023, Sal Khan proclaimed, “We’re at the cusp of using AI for probably the biggest positive transformation that education has ever seen. And the way we are going to do that is by giving every student on the planet an artificially intelligent but amazing personal tutor.” Same dream, different era, but with one key difference: Who is building the tools, and why?

In the 20th century, using technology to create access to education was a federally funded research project based on the latest science on learning and development. Today, the companies developing AI tools to enhance learning are driven by financial returns. What do we risk when we rely on for-profit companies to develop education technologies? What can we learn from the ed-tech innovators of the previous century that can help us design a future where education gives everyone the chance to enhance their potential?

Throughout the 1960s, three institutions—Stanford, the Massachusetts Institute of Technology and the University of Illinois—backed significant experiments in using computers in education. Stanford’s Patrick Suppes pursued the “individual Aristotle,” which became the model for intelligent tutors and instructional chat bots and created adaptive learning. MIT’s Seymour Papert, an inaugural co-director (with Marvin Minsky) of the Artificial Intelligence Lab (now CSAIL), created a programming language for children so that they might use computers as a tool for learning. And Donald Bitzer, of the University of Illinois, created PLATO, a networked course distribution system that educated up to 1,000 learners at once, featuring touch-screen technology, plasma displays and real-time communication with other users.

The creative visions of these three men are embedded in almost every major technology deployed for education, from touch screens and toy robots to chat bots (e.g. Google’s Bard and ChatGPT) and adaptive instructional programs to online courses. Yet their history has been buried under a constant stream of educational innovations that claim to solve “new” educational problems that we have been addressing for nearly a century.

Four lessons from early ed tech still apply. First, we should recall Papert’s idea to resist “the computer … being used to program the child” and instead create opportunities wherein “the child programs the computer.” Technology is a tool, and learners benefit when they understand its structure, not just its outputs. Second, AI-powered tutors should be used to deepen understanding and allow faculty to focus on higher-value interactions. Third, technology should be used to foster connections between learners to make learning more social and contextual. And finally, technology in education shouldn’t seek to make learning more cost-efficient or scalable, but instead make it more engaging and meaningful.

These ideas are especially urgent as generative AI raises profound questions for higher education. Many of the AI tools currently celebrated as transformative reflect narrow, utilitarian views of learning. They promise automation, personalization and skill acquisition but sideline more valuable aspects like teamwork, critical thinking and problem-solving.

Universities have a necessary role to play in shaping how AI supports learning. The technology has advanced sufficiently to fulfill aspects of the early pioneers’ visions: Suppes’s adaptive tutor, Bitzer’s social learning network and Papert’s technology-enabled discovery. But to realize that potential, universities must reframe the challenge AI poses not as an existential threat, but as a symptom of deeper structural problems: inequity, commercialization and reductive definitions of learning. And universities should lead in resisting the fantasy that education and learning should be efficient and scalable. Learning is often messy, slow, recursive and profoundly inefficient. Today, those who call for efficiency in education often stand to profit from the solution.

Higher ed is at a curious moment, staring down twin barrels of a potentially field-changing technology and intensifying political pressure. We can learn from the experiments of the 1960s and ’70s that designing for deeper intellectual agency, broader access and learner empowerment is often at odds with designing for profitability. Universities have a once-in-a-lifetime opportunity to use these powerful new tools not to remake education in the image of the market but to restore its purpose as a public, human and shared endeavor.

Using Papert’s idea that technology should serve human behavior and not reprogram it, higher education can use AI tools to support an emerging new model of education, one that is affordable, blended, lifelong and flexible. Designed well, this model could reduce the barriers that prevent Americans from moving between education and work throughout their lives. College-level education should double down on its commitment to use AI to learn how students learn while teaching students how to be learners throughout life.

As the pace of new technology accelerates, careers will shift more frequently, and adults will re-enter education repeatedly. Four-year degrees will remain vital and transformative for many, but they will be just one credential in a wider, more dynamic system. High schools and businesses will also offer learning experiences that can be recognized with a meaningful credential. The future of education should be defined not only by how well it prepares us to be workers but how well it prepares us to be active participants in society. We should be teaching students not how to be slaves to the machine, but shapers of its purpose.