ILLUSTRATION: NHUNG LE
People were understandably intrigued by this powerful new tool—the highest-profile exemplar of so-called generative AI—which can, within a few seconds and after a few prompts, create text, images, music, and computer programs.
Even before ChatGPT became part of the popular discourse, MIT was looking to the future of AI. Building on that experience, the Institute provided exploratory funding in the fall of 2023 for more than two dozen research proposals aimed at determining how this new technology might be utilized for societal good and identifying potential pitfalls to be avoided. Four MIT researchers—Hal Abelson PhD ’73, the Class of 1922 Professor of computer science and engineering; Cynthia Breazeal SM ’93, ScD ’00, MIT dean for digital learning and professor of media arts and sciences; Eric Klopfer, professor and director of the Scheller Teacher Education Program and The Education Arcade at MIT; and Justin Reich, associate professor of digital media in MIT Comparative Media Studies/Writing and director of the Teaching Systems Lab—received a grant to explore the implications of generative AI on K–12 education.
Fear of the unknown
“A lot of the initial perspectives focused on worries about students using generative AI to cheat on papers,” says Klopfer, who leads the MIT Comparative Media Studies/Writing program. Although that concern must be addressed, he maintains, “we are trying to change that conversation. Schools are not walled gardens separate from the rest of the world. Since the rest of the world has access to these tools, students need to know how to use these tools effectively.”
“Banning technology doesn’t work well,” adds Reich. “It just encourages students to figure out ways of getting around those bans.” Nevertheless, there are ways of walling off technology, at least temporarily. Teachers, for instance, can prevent students from using calculators in class until they establish proficiency in arithmetic. Once they’ve done so, students can use calculators to free up time for more intellectually challenging math problems.
Writing assignments can teach students to not only write but also to think. “So we need to be careful about introducing technology that enables students to bypass the kind of thinking we want them to do,” Reich says. The challenge for teachers, he observes, is to figure out what things we should let ChatGPT do and what things we should prevent it from doing.
New technology, familiar challenges
This dilemma is by no means novel. Schools had to devise policies governing the use of calculators, indicating when their use is permissible and even desirable. With the internet widely accessible in US schools for the last 25 years, educators like Reich, who taught high school history from 2003 to 2007, worried about whether students were cutting and pasting material from online into their papers. Yet this technology also brought obvious advantages: students could quickly gather information from books, newspapers, and journals, providing more time for reading and processing information, rather than rummaging through library stacks or poring over microfilm.
The hope is that generative AI can—if properly applied—offer benefits to K–12 education that greatly outweigh its potential downside. In the road map the MIT researchers have prepared for schools, they have spelled out what a “brighter future” for AI in education could look like, detailing the key steps needed to achieve that vision.
“Kids learn by creating things,” says Klopfer, and generative AI can help them do that in myriad ways. He points to Aptly, a software platform recently developed by Abelson along with an MIT software engineer and two undergraduates, “as representative of the kind of things we want to point children toward.” Aptly makes the chore of designing apps for a smartphone easy. “You can speak to your browser and say: ‘Make me a smartphone app with a few buttons, each corresponding to a different language,’” Abelson explains. “When I press a button, translate what I speak to that language.”
The range of apps you can make is essentially unlimited, Abelson says. “If a computer can make any kind of program you want, we then have to decide what to teach our kids in terms of computing. And once you make what you wanted, how would that benefit the world?”