After I graduated, I used my first paycheck to buy a telescope. That was the first time I tried to learn something more seriously without anyone asking me to. I quickly realized that I didn’t understand much, and even basic things felt difficult at the beginning.

What pushed me to keep going was a very simple reason: there were things I wanted to see, but I couldn’t see them unless I learned how. At one point, I wanted to observe a comet that was barely visible, and I realized that if I didn’t figure out new ways to do it, I would simply miss it. That feeling made me willing to sit with confusion and keep trying, even when progress was slow.

The first time I tried, it didn’t go very well. But going through that experience made a difference. The next time, I felt more comfortable experimenting and trying different approaches. I was less worried about not understanding everything and more willing to explore. Once I had some experience, I also became more confident and more motivated to continue. This changed how I think about learning. I don’t think having the “right background” is always necessary to start. In many cases, I only begin to understand something after I try it myself.

Outside of astronomy, I also enjoy making things and experimenting with different materials. I don’t have any formal training in art or design, and at first it is easy to feel like I am not capable of doing it well. But I’ve found that it helps to start small, try things out, and adjust along the way.

I also realized that it is not always helpful to only look at experts. Sometimes it is easier to learn from people who are still figuring things out, because their process feels closer to mine. Seeing how others approach things at different levels helps me find my own pace.

I think this is what makes me “sticky” to learning. It usually starts from something small that I care about, but once I notice the gap between what I want and what I can do, I feel motivated to close that gap. I don’t always understand things at first, but the process of gradually figuring things out becomes rewarding on its own.

This pattern also shows up in my previous background in computer science and in my current work in education. I tend to stay with something not because I have to, but because I want to understand it better. Once I start, it is hard for me to stop halfway.

Jiayue Chai, M.S.Ed, LST '26

When does play emerge in human development? This question continues to fascinate me. I grew up in El Salvador during a period of violence, where gangs ruled my neighborhood. To keep me safe, my parents did everything in their power to keep me indoors. Inadvertently, crafting a world of discovery within our home. By providing access to LEGO bricks, jigsaw puzzles, and video games, they shaped the lens through which I view learning today.

Because my father was an electrician and my mother a secretary, I had a typewriter at my disposal and an overabundance of wires and tools to explore. In my imagination, these components became spaceships and obstacles for my characters. These early exposures influenced me to think of technology not just as hardware, but as parts of larger systems. They were experiences that have the power to shape the world.

That is how I view education today: the role of an educator is to craft experiences. Learning transcends the textbook or the laptop; it taps into the classroom environment, the school community, and the national policy framework. This is the root of my fascination with games as pedagogical tools. Games are systems of perpetual learning where player agency and experimentation drive progress, in a world made by rules (policy). I am driven to uncover the lessons we can extract from these systems of play and apply them to the rigorous demands of modern education.

When Games Became Learning

Directly after high school, I got an internship creating 3D models for oil companies, a skill I had cultivated during high school. As the World Wide Web transitioned to Web 2.0 and Flash began to phase out, I was tasked with building interactive simulations using the Unity game engine. My job was to collaborate with engineers in the energy sector to translate complex technical requirements into VR spaces and interactive games for training a younger workforce.

This experience led to a collaboration with Dr. Hannah Gerber at Sam Houston State University, where we developed a mobile game focused on pro-social outcomes, specifically, recycling. We observed how youth in Egypt took the lessons from the game and initiated actual recycling programs in their villages. These results cemented my belief that games are profoundly conducive to learning.

When I transitioned into teaching, I had the opportunity to develop a Technology and Design curriculum for students from second through twelfth grade. I approached my classes as if they were a game, applying design thinking across various developmental stages and subjects. I eventually adapted these games and learning pedagogy for International Baccalaureate, Montessori, and Special Education environments. Witnessing students flourish when granted the agency inherent in games remains one of my most impactful experiences. There is something about allowing students to have this agency in the classroom, that triggers a wondering in them, a curiosity, and I have watched those early embers of curiosity evolve into full career paths and college decisions years later.

The New Character in the Game: AI

I first integrated AI into my teaching when I was working at a school for students with language-based learning disabilities. For these students, reading and writing were often insurmountable barriers. Watching them develop a desire for reading and writing because they could speak their ideas to an AI to gain agency over those barriers was transformative for both them and me.

A few months after, I began my Master’s at the Harvard Graduate School of Education during the first academic year that Generative AI was available. I witnessed firsthand the friction between administrators, faculty, and students because of AI. Fortunately, I got involved doing a research study interviewing my peers about their AI usage. The difference in performance between those that used AI regardless of the school policy, and those that did not want to use AI due to fears of retaliation from faculty, was stark. It became very obvious to me that AI was an equity issue in education and would generate even wider gaps in education than we have today.

In response, I began organizing AI workshops in Cambridge, Massachusetts. With the support of Professor Seiji Isotani, I extended these workshops to the Ministry of Education in El Salvador to help implement national policies for AI literacy. I have since taken this work to universities in Japan and India, and high schools in China. With data from three years of doing these workshops and gathering student reflections and observations, I developed the “Stages of Learning AI” framework, which outlines the learning arc students follow when introduced to Generative AI tools. This journey has also led me to do experiments in utilizing AI avatars within VR for language learning in partnership with Dr. Takako Aikawa at the MIT Global Languages department. A theme I still continue to explore as a Visiting Researcher at Reitaku University. As well as exploring the ethics of human digital twins, the ownership of data, and pro-social behaviors mediated by play at the Morality Lab at Boston College.

AIED at GSE: The Future of Learning

AI is undeniably redefining the educational experience. Its integration has highlighted the deficiencies in our current systems while revealing the work yet to be done in the learning sciences and policy to ensure equitable, quality, education for all.

AI also lowers the barrier for educators to bring games and simulations into their own practice through vibe-coding and rapid prototyping. I am deeply interested in helping teachers master these tools while studying how pedagogy must evolve to meet these new capabilities. Furthermore, I am exploring the relationship between motivation and agency: In a world where intelligence is now accessible to anyone, how can an AI encourage a student to remain an agent of their own growth? And what is the importance of play as a learning tool in this new age of technological advancement? Can the intersection of play and AI help with education at scale? These are the questions and experiments I look forward to exploring with colleagues at the AIED Lab.

So who am I?

I am Beryl. A Kenyan teacher at heart. A person who feels strangely calm in front of a computer screen. Someone who finds peace in the quiet rhythm of typing, even when the code is wrong, even when the logic breaks somewhere in between.

A good day for me is a day where I am building something. A bad day is when I feel unstructured. And I LOVE structure. And yet, lately, life has been gently teaching me how to live without it.

For a long time, I was “just” a teacher. I taught mathematics and business studies in Kenya, carrying chalk, lesson plans, and a relentless belief in my students, sometimes stronger than the belief they had in themselves. Teaching was never just technical for me, and it is as much relational. It has never about regurgitating new information but to experience the closeness I share with my students.

I watched students navigate pressures far beyond the classroom.
I watched their potentials got bend under circumstance.

And as teachers during these times, the most human thing we could do is to try to hold space.

Somewhere along the way, the questions began to grow bigger for me.
I used to constantly asked myself “How do I teach this better?” but as the questions grow, I can’t help but wondering Why do we teachers notice struggles so late, What patterns are we missing? What if we could recognise quiet struggle before it becomes visible failure?

These questions led me back to the computer. There is something grounding about computing for me. When I open my laptop, I feel clear, structured, predictable. Even when I don’t have answers, I can still somewhat trust the process. There is a clear logic beneath the surface holding everything together.

Studying computing is teaching me to see systems. Data flows, architectures, invisible scaffolding. At the same time, studying education keeps me tender. It reminds me that learners are not variables and I learned to view and treasured their lived experience and not simply treat them as datasets.

Being part of the AIED Lab reminds me of what it feels like standing in that in-between space. It is a place where algorithms meet classrooms. Where curiosity is shared and where questions about AI are not just technical, but deeply human.

I am drawn to AI in education not because it is futuristic, but because it is fragile but perhaps because it invites gentler questions. How do we design systems that support teachers rather than replace them? How do we handle behavioural data with care?
and most importantly: How do we build intelligence without unintentionally widening inequities?

Growing up in Kisumu shaped my sensitivity to inequality. Teaching made me empathetic. Computing is making me precise. Some days, I miss the certainty of being “just” a teacher. That identity was clear. Defined. Held. Now, I feel more layered. Less certain. More open. Still kind and structured, but just at a different scale.

So who am I?

I am Beryl.
A former classroom teacher.
A computing student.
An education researcher in the making.
Someone learning how to hold code and classrooms in the same hands.

And maybe this season is not about being lost.
Maybe it is about becoming. Slowly. Intentionally.
Learning to build the future of learning, one idea, one system, one small question at a time.

Beryl Odingo

Editorial: Besides pursuing her two Masters in Education Studies and Computer Science at the same time, Beryl also serves as a Lead Research Apprentice of Community Learning in our Lab.

My tryst with technology in education started early—one half of my parents being a teacher and artist, the other an engineer fascinated by electrical and electronic systems. Conversations at home often moved fluidly between creativity, teaching, and how systems work. Looking back, it feels inevitable that I would eventually become interested in how technology and learning intersect.

I also grew up watching education transform in real time. Classrooms shifted from blackboards and chalk to projectors and smart screens. The internet boom made information more accessible, and slowly, the boundaries of the classroom began to expand.

For me, that shift became personal during my teenage years. I was learning trigonometry from Sal Khan videos. Those videos were simple, patient, and available whenever I needed them. Long before we talked about “AI tutors,” millions of students were already learning from the internet. At the time, it didn’t feel revolutionary—it was just a helpful explanation that made learning click.

The Crossing: Lessons from the Classroom Floor

Years later, those early experiences began to connect with my professional life. After working in government schools in India and later working with undergraduate engineering students,I realized that technology alone doesn’t automatically improve learning. Some tools open doors; others create barriers. Access varies widely, and tools designed in one context don’t always translate well into another.

Those experiences pushed me to think about broader questions: How does technology shape learning? Who benefits, and who gets left out?

Today, as a doctoral student in the Learning Sciences and Technologies program at the University of Pennsylvania Graduate School of Education, these questions guide my research. I explore how artificial intelligence—particularly large language models—can support learning at scale, especially in contexts where educational resources are limited. The Penn AI Fellowship supports this work, providing mentorship, resources, and an interdisciplinary community where AI can be studied responsibly and inclusively.

Much of my research draws on massive open online courses (MOOCs), where tens of thousands of learners can enroll in a single class. These environments offer an extraordinary opportunity to understand learning at scale, but they also create challenges: instructors cannot interact individually with every student.

To understand learning in these spaces, I study the digital traces students leave behind—discussion posts, votes, time on assignments, and patterns of participation. These signals reveal how learners engage with course material and with each other, and how those behaviors relate to outcomes like course completion, grades, or professional advancement.

At the same time, I’ve been exploring how AI can support educational research itself. In one project, I used GPT-based models to analyze thousands of student discussion posts. Instead of manually coding every comment, we experimented with using AI to help identify themes in what students were saying. The goal was not to replace human researchers, but to understand when AI can assist with large-scale analysis and when human interpretation remains essential.

More recently, my work has focused on a more fundamental question: can AI meaningfully explain things to learners?

The Inquiry: Will AI Expand or Limit Opportunity?

Working with Prof Isotani and colleagues at the AI in Education Lab, I study how well large language models provide explanations, offer feedback, and adapt to learners with different backgrounds or goals. In theory, these systems could make learning more personalized and accessible. But many are trained primarily on English-language data and Western contexts. If AI tools are to support learners globally, we need to understand whether their explanations resonate with students from different cultures, educational systems, and resource environments.

This question feels especially urgent given my experiences in classrooms in India. Education technologies often promise to democratize learning, but tools designed in one context do not always translate well to another.

Looking ahead, my research is increasingly focused on how AI-based learning tools can work in diverse and low-resource contexts. This includes studying how learners interpret AI-generated explanations, exploring smaller and more deployable AI models, and thinking about how these tools can support—not replace—teachers.

In many ways, the questions guiding my work today trace back to those early moments: growing up in a household where teaching and engineering coexisted, watching classrooms digitize, and learning trigonometry from online videos.

We are entering another major shift. Artificial intelligence is beginning to reshape how knowledge is created, shared, and learned. The challenge is not simply building more powerful systems—it is ensuring these technologies expand educational opportunity across cultures, contexts, and communities.

The Penn AI Fellowship provides a platform to explore these questions alongside scholars from diverse fields, creating the mentorship, community, and resources to think about AI not just as a tool, but as a force that can be responsibly and inclusively shaped for education worldwide.

Shruti Mehta

It began in early September. We were just a few weeks into the new semester at Penn GSE when an email arrived from the Division of Learning, Teaching, and Leadership. It was an invitation to participate in the Faculty Research Apprenticeships program—a chance to open our current projects to students looking for credit-bearing research experience.

The “3-to-5” Rule

Being researchers, we started with the data. Looking at postings from previous years, the pattern was clear: most faculty members took on between three and five students per semester. It made sense; these roles require heavy mentorship and deep one-on-one engagement.

With that “3-to-5” range in mind, the two of us (Seiji and Nam) sat down and mapped out a few projects we currently had in motion. We expected a small, focused group of students from the Learning Sciences and Technology (LST) division to join us.

A Surprise in the Inbox

We couldn’t have been more wrong.

Instead of a handful of applicants, we saw an overwhelming flood of interest. But it wasn’t just the volume that surprised us—it was the breadth. Applications weren’t just coming from LST; they were pouring in from across the entire school where the Apprenticeship is not available

• International Education Development

• Education Policy

• TESOL (Teaching English to Speakers of Other Languages)

• Education Leadership

• Statistics and Measurement

The Ethnographer’s Approach

Faced with this sudden community of 38 eager students, we did what any researcher does when entering a new environment: we listened & observed.

We spent time learning about every single one of those 38 students. We wanted to understand their “where” and their “why.” What we found were 38 fiercely independent thinkers with massive potential and endless curiosity.

We noticed a disconnect in the current landscape: AI in the industry—and even as a research practice—draws immense attention, but it rarely invites participation. It often feels like a closed door. We decided then to make a bold move: we would create (or invented) new “problem spaces” of our own and invite our students to help us build it.

Defining the Vision

For three weeks, we worked to turn sketches into a concrete vision. We centered our work on two pressing questions:

1. Human-Centered AI: What are the big generative topics in instruction and teaching that allow for technology adoption without losing the human element?

2. Community of Practice: How do we build a structure where people support and learn from each other, rather than just from a professor?

The “Thanksgiving Test”

Finally, it was time to put the motivation of this new group to the test.

We scheduled our very first AI lab meeting for a Wednesday at 4:00 PM—the day right before Thanksgiving break. On a campus that is usually emptying out for the holiday, we weren’t sure what to expect.

To our absolute surprise, every single student who RSVP’d showed up.

Every. Single. One.

• Some took a 45-minute bus ride to get there.

• Some traveled for two hours through the November chill.

Standing in that room, seeing the dedication of these students, gave us the hope and fuel we needed. It proved that this space was necessary. This wasn’t just a research project anymore; it was a movement.

This is where our story begins.

Subscribe now