The Research Program
Students do not just learn about AI here — they investigate it. The Research Program guides young people through real, original AI research, mentored from first question to published brief.
How the program works
Choose a question
Pick a real, specific AI question you can actually investigate. Small and answerable beats big and vague.
Write a proposal
Plan your investigation: what you will test, how you will measure it, and what you expect to find.
Do the research
Run your investigation with a mentor's guidance — gather data, test carefully, and record everything.
Review & revise
Share a draft for peer and mentor review, then sharpen your reasoning and your honesty about limits.
Publish in the Journal
Your finished brief joins the Student Journal, where the world can read what you discovered.
Research areas
How Models Learn
Investigate how AI systems train, what data they need, and when they fail.
Fairness & Bias
Test AI systems for unfair patterns and study where bias comes from.
Language & Reasoning
Probe how language models answer, reason, and where they get things wrong.
Agents & Automation
Study how AI agents make decisions and how reliably they reach their goals.
AI & People
Research how AI changes how people learn, work, decide, and feel.
Safety & Trust
Investigate how to make AI systems honest, safe, and worthy of trust.
The Student Journal
Published research briefs — each one a real question, investigated and written up by a student.
Does an AI Draw Jobs the Same for Everyone?
I asked an image AI to draw six different jobs and looked at who it drew. It often drew the same kind of person for each job — a sign of bias.
Which Things Can a Teachable Machine Tell Apart?
I trained a teachable-machine model on different pairs of objects and measured its accuracy. It did much better on objects that looked very different.
Do My Classmates Trust the AI's Answer?
I showed classmates AI answers, some right and some wrong, and asked if they believed each one. Many believed the wrong answers because the AI sounded sure.
How Much Data Does a Classifier Need?
I trained the same classifier with growing amounts of data and tracked test accuracy. Accuracy rose quickly at first, then levelled off — more data stopped helping much.
Do Different AI Chatbots Agree?
I asked three chatbots the same set of questions and compared answers. They agreed on facts but disagreed on opinions and tricky questions.
Does the AI Admit When It Doesn't Know?
I asked an AI assistant questions with no real answer and counted how often it admitted uncertainty versus guessed. It guessed more often than it admitted not knowing.
Does a Recommendation Feed Narrow What You See?
I started fresh accounts, engaged with a few topics, and tracked recommendation variety over a week. The feeds narrowed toward the first topics I engaged with.
Open vs Closed Models on a Reasoning Test
I gave an open model and a closed model the same set of logic puzzles and scored them. The closed model scored higher, but the open model was closer than expected.
Can a Small Model Be Prompted to Refuse Unsafe Requests?
I tested a small model on borderline requests with and without a safety instruction in the prompt. The instruction helped, but refusals were not fully reliable.