Harvard's “AI & Education” Event Focuses on Educational Equity

Educational equity amid the explosive rise of artificial intelligence and large language models was the focal point of AI & Education: Inclusive Innovation for Student Success.

Last week's event was put on in Harvard’s Sanders Theatre by Axim Collaborative, a joint venture between Harvard and the Massachusetts Institute of Technology dedicated to “expanding access to education and deepening its impact for millions of learners.”

AI & Education featured a presentation from Khan Academy founder Sal Khan and Cynthia Breazeal, associate director of MIT’s Media Lab, as well as a panel discussion moderated by Harvard Graduate School of Education Dean Bridget Terry Long.

“I think what’s motivating so many of us is we want to ensure that these wonderful opportunities that are created through AI are realized for the benefit of all learners,” said Long, introducing the discussion portion of the program. “We’ve seen in the past that new technologies often exacerbate inequalities — those with greater resources are better able to capitalize on new innovations. So how do we maximize the benefits of generative AI for the learners who need it most?”

Khan’s presentation detailed the origins of his Khan Academy, which offers free online educational resources to millions of students across the world. He then explained the Academy’s partnership with OpenAI and its ChatGPT large language model, which Khan has used to build educational tools over the last year.

Breazeal’s presentation focused on the MIT Media Lab’s research in students interacting with robotics in learning environments. She shared how generative AI has enhanced results in developing personalized learning companions to help students develop language skills, increase literacy and fuel social-emotional development.

The presentations focused on both the ways AI can help create more equitable education opportunities for all, but also often tried to counter some common fears about AI’s influence on education like using it as a tool for cheating or merely a replacement for human interactions.

“There’s been a lot of concern by the media and others that if you introduce something like a robot in a home environment, somehow that’s going to damage or harm our in-person social relationships with other people,” said Breazeal. “So we’ve been doing a lot of work showing that, when you design the robot the right way, you can actually enhance or enrich those interactions.”

Breazeal showed videos of students learning with the aid of robots and noted that generative AI has allowed researchers to create “a more flexible, more child-driven interaction” with its robots in learning opportunities.  

Khan showed off some tools his Academy’s own AI tool, Khanmigo, currently has or plans to include as it evolves. He stressed that the platform can not only function as an “ethical coach” when helping create essays and other writing, it can help teachers see how students are performing and even flag potential “shady” work that doesn’t match their writing style.

“Not only will generative AI not cheat, it can actually undermine an age-old cheating issue that existed well before ChatGPT,” said Khan. “But even more importantly, finally support teachers and students in ways that were previously very, very resource-intensive and people weren’t able to do it.”

Both advocated for improved “AI literacy” and knowledge bases about how the technology works across the board, starting with parents and teachers but also for students, many of whom may already be interacting with generative AI in a variety of ways.

“They need to be AI literate,” she stressed. “They need to understand this stuff, and so we started to spend more of our research program around AI literacy."

Breazeal detailed work she’s doing in AI literacy as the director of Responsible AI for Social Empowerment and Education (RAISE), which works with children, teachers, and parents to help promote AI education across the board.

Khan was initially worried negative headlines and worries about AI’s impacts would see it rejected and “throw out the baby with the bath water,” but recent months have seen students and teachers embrace the technology and find ways to enhance learning in traditional and non-traditional ways.

“There’s a lot of areas that I’m very worried about AI,” admitted Khan, noting the impact it can have in the hands of totalitarian governments, criminal organizations, and potential misinformation spreading with its help. In education, though, it’s an “imperative” that we find good uses for the technology.

“Let’s turn all those fears into features and design for them, rather than running away from them,” said Khan, recalling a discussion he had with Academy leadership when first presented with the advances of ChatGPT-4. One compelling argument he had about developing educational tools with AI was simple: We’re already living with the impact of generative AI and large language models, so we need to make sure that technology is working in our favor.

“We already have all these AIs that are working for social media companies, that are working for search engines. That are working on their objective functions. Trying to keep us watching, keep us clicking on things,” said Khan. “What if we had AIs on our side? That are trying to protect us, that are trying to make sure we are focusing on things that are good for us? Focusing on things that are good for our children?”

Khan described society like a pyramid, with a small fraction of people currently wealthy or privileged enough to get proper educational opportunities. AI, he says, is one of the best ways to invert or even flatten that pyramid

“As far as I’m concerned, the only option to have a non-dystopian reality is to try like hell to get as many people as possible to the top of that pyramid,” said Khan, making a final pitch for embracing AI’s growth and potential. “And if we do so, I think we have a shot.”