There are some moments that you will always remember. When something clicked, usually because someone said the right thing at the right time. Not when you passed a test or finished a course. A teacher who saw that you were lost. A stranger in a web forum who went on and on about something they didn’t need to explain for ten minutes. This social, sometimes unpredictable way of learning is exactly what Caitlin Morris wants to take care of.
Morris is a MAD Fellow and a doctoral student at MIT‘s Media Lab. Her research is in the awkward area between what AI is really good at and what it consistently misses. She grew up in a rural area of New York, far from the kinds of schools that would have normally helped a child become a technologist. She started learning how to code in online communities instead of classrooms. She read through threads, worked backwards on public projects, and slowly figured things out with people she’d never meet in person. “That was this huge wake-up moment of feeling like there was a path to expression that wasn’t a traditional computer-science classroom,” she has said of that time. This is the kind of origin story that makes a researcher care about certain things. And for Morris, it changed everything.
The Fluid Interfaces Group at the Media Lab is where she works on a question that seems simple at first but is actually quite hard to answer: what parts of human interaction in a learning setting really can’t be replaced by AI? Not what AI does badly right now—technology gets better all the time—but what is fundamentally and almost philosophically unique about learning with other people. It’s possible that no one has a clear answer yet. Morris doesn’t seem to mind that she doesn’t know, which is likely why she’s setting up ways to find out instead of assuming she already does.
The tools she’s making will let her see both behavior and feelings at the same time. Along with reflections from the learners themselves, which were gathered through interviews and structured self-reporting, things like body language, language patterns, and the way someone hesitates or leans in are all taken into account. The goal is to link what people do with how they feel while they do it, which is something that most edtech platforms don’t do. Most platforms track how many quizzes are taken and how many are completed. Morris is looking at more than just how much a student feels like they belong.
She co-organized the MIT Media Lab’s Festival of Learning and ran events for creative coding communities for years before she got her PhD. She has also taught technology and design in high school and college. Because she has been an organizer, a teacher, and a researcher, her work has a grounded quality that pure academic research doesn’t always have. She brings what she’s learned from seeing real students struggle and succeed in real rooms into the lab.
Some people in the field of educational technology think that AI can solve the problem of scalability in learning by making it possible to copy one good system over and over again. And I think it can, up to a point. But Morris’s research is quietly looking into what is lost when things are sped up. She says that social dynamics have a direct effect on curiosity and what drives people on their own. We don’t know yet if these dynamics can survive being translated into AI-supported environments. It’s still not clear if even the best-designed tools can replicate what happens when a peer asks you a question that changes the way you think about a problem.
MIT is at the center of the larger conversation about AI in education. As part of its Open Learning initiative, researchers are looking into how these tools should be responsibly made, regulated, and used. Morris adds something specific and, it seems, necessary to the conversation: he keeps reminding us that efficiency isn’t the same as education and that a student’s motivation might depend on whether they feel like someone sees them and genuinely cares about how they’re doing. No matter how smart AI is, it hasn’t figured that one out yet.
