How University of Peradeniya evaluated conceptual mastery.
Axiom Flow served as a graded assessment layer for 100 students in Computer Systems Programming, seamlessly integrated via Moodle. Unlike traditional quizzes that test simple recall, Axiom Flow evaluates conceptual mastery by having students teach Sam, an AI model that only knows what students explicitly teach him. Here is what 52 students said about their experience.
The Setup
100
Active Cohort
A cohort of 100 active students in Computer Systems Programming.
LMS
Moodle Integration
Seamless single sign-on and integration with Moodle LMS.
Sync
Automated Grades
Direct, automated grading synchronization with the Moodle gradebook.
52
Survey Responses
Students completed a comprehensive post-session feedback survey.
Key Metrics from 52 Responses
86.5%
“Did teaching Sam help you understand the topic better than studying alone?”
Answered 'Yes'.
86.5%
“How did you feel about Sam's questions during the teaching phase?”
Answered 'About right'.
88.4%
“How does Axiom Flow compare to a regular quiz or assignment?”
Answered 'Better' or 'Much better'.
61.5%
“Did the final score feel like a fair reflection of how well you understood the topic?”
Answered 'Yes' (with another 30.8% answering 'Not sure').
In Their Own Words
“Sam used to ask really logical questions when we gave partially correct answers, which directs us to go find the real accurate descriptions and come back and teach Sam while learning them to the point ourselves.”
“I love teaching Sam and how it surprises me pointing out the things that even I didn't know that I don't know about the topic.”
“It was easy to catch my misconceptions.”
“I learned many things because I had to teach it. I enjoyed learning like that.”
“I feel I'm a really good teacher when Sam is facing the exam alone.”
Lessons Learned & What's Next
Axiom Flow is a rapidly evolving platform. Transparently tracking where the product has room to grow is essential to our iterative approach. Here is how we are turning cohort feedback into active development goals.
Context Memory Expansion
“Some students felt Sam lost context across multiple, extended assignments.”
We are implementing deeper conversational memory architectures to ensure Sam retains previously explained concepts seamlessly across sessions.
Tuning the Prompt Baseline
“A few students noted Sam asked too many basic clarifying questions.”
We are adjusting the AI's baseline knowledge threshold so it focuses on core logic rather than getting hung up on basic syntax.
Clarifying AI Communication
“Sam's responses could occasionally be difficult to parse when correcting students.”
We are refining the generation layer to be more concise and explicit about exactly which part of the student's answer is missing.
Progress Transparency
“Students were frustrated when progress plateaued and it wasn't clear what specific concept Sam was still missing.”
We are building clearer 'knowledge gap' hints that nudge students toward the missing sub-topics without giving away the exact answer.
The Takeaway
In a live cohort of 100 students, the results were clear: teaching Sam drove better understanding than studying alone, and students vastly preferred it over regular quizzes. The primary focus for our next iteration is refining our grading algorithms to ensure Sam's scoring consistently feels fair and precise, and addressing friction points around Sam's context retention.
Bring authentic, teaching-based assessment to your curriculum.
Book a brief walkthrough to see how Axiom Flow, an AI-powered assessment platform, integrates with your LMS to verify conceptual understanding through the Feynman Technique.
Book a Demo