The Learning Engineering Virtual Institute, or LEVI, is a collaboration of researchers, learning engineers, and educators striving to develop, scale, and implement new tutoring platforms that can double the rate of math progress among middle school students, especially those from low-income backgrounds. Since 2022, seven teams have been working to achieve this goal. Their ideas range from an AI-powered chatbot that provides personalized math tutoring to AI video technology that digitally replicates the experience of having a personal tutoring session.
Working with schools in the UK, the team from Eedi is building a “misconception map” that links to insights and analytics about student math misconceptions and suggested interventions for teachers. A “misconception” is a fundamental misunderstanding of a mathematical concept that leads to consistent errors, no matter how strong a student’s computation skills are. Identifying and addressing misconceptions is key to building confidence and improving math performance.
The Eedi team has built a learning platform that can identify and resolve over 90 percent of students’ misconceptions using best-in-class pedagogy and tutors. The platform uses AI to help unpack students’ misconceptions and helps predict future misconceptions if the misconception is not resolved.
In this installment of “5 Questions,” Ben Caulfield, Eedi’s CEO, discusses the team’s work and findings.
What was your “ah-ha” moment when you knew you were onto something workable?
Understanding the true value of the assessment data we’d accumulated over the last 10 years. Not the fiscal value, but rather the value of transforming education.
We chose to focus on misconceptions from the outset, not to just identify knowledge gaps, but to understand why they exist. This uniquely powerful dataset gave us insights into how misconceptions evolved over time, but when you combine that with machine learning, magic happens.
That “ah-ha” moment was the realization that we could predict how a student would perform against 2,000 math skills from just 10 questions with a high degree of confidence. Those results gave us hope. Those predictions are not set in stone, any intervention can help improve learning outcomes, but when you provide those insights to the teacher in the class, the parent at home, a tutor, the child, publishers, or a government, we can change the future.
Have you made any significant shifts or course corrections?
We embrace our values. Always be learning. Promote evidence. We have applied existing research to our solution, run robust A/B experiments to validate our assumptions, and shared the results widely.
Another value: Be collectively responsible. We’re a small but mighty team full of amazing talent but we recognise that we don’t know everything. We originally wanted to take a causal discovery approach to learning a misconception map but, when it seemed this might not work for the number of misconceptions, we were steered towards a more scalable link prediction methodology by LEVI’s knowledgeable advisors.
The most significant change, however, has been driven by impatience – not a Company value! We are building cutting-edge models that not only predict knowledge gaps but will predict the likely misconceptions that will hold a student back. We want to get those insights into the hands of people who can act: teachers, students, parents, tutors, government.
We’ve not cracked growth, few do, and even fewer balance growth with impact and profit, but a shift to work with partners with existing traction means we are closer than ever to achieving our mission to help improve learning outcomes for 1 billion children by 2030. Whether it’s powering the world’s best AI tutor, a tutor co-pilot, or an innovative approach to assessment, we believe Eedi’s work with LEVI will change lives.
What’s been the most surprising thing your team has learned?
The capital required to make any sort of dent in the U.S. market.
From an educational standpoint, there is very little that unites the 50 states and that gulf continues to grow, making it increasingly difficult for providers. States and districts are dictating more, making it harder for teachers to adopt solutions that they feel are right for them and work.
Never say never, but gone are the days when a bottom-up sales approach, adopted successfully by the likes of Kahoot! and ClassDojo, will deliver the same results. A top-down approach requires feet on the ground and RFP writers. Start-ups generally don’t have the capital nor the time.
When did you see Eedi working for students and teachers?
Wednesday, Sept. 25, 2024.
Eedi’s Director of Learning Sciences, Bibi Groot, delivered the early findings of a year-long random control trial involving roughly 3,000 11-year-olds. Independently evaluated by What Worked Education using NWEA’s MAP assessment, one graph stood out amongst a myriad of incredible results.
A single graph highlighted the learning gains of low-income students against dosage. The disadvantage gap unhappily separated the two lines but at a dosage of 120 questions, something amazing happened. Low-income students started demonstrating sizable learning gains compared to their better-off peers. At 200 questions, the gap in score improvement stopped widening, and we began to close it. Incredible.
Laurence Holt noted in “The 5 Percent Problem,” that many math programs may most benefit the kids who need it least. The results of Eedi’s RCT highlighted that over 50 percent of students achieved the required dosage of 120 questions. Doubly incredible.
Learning and teaching is a human endeavor and all of these results were highlighted in a single moment by Sam, a Learning Designer and ex-teacher working at Eedi. Seeing that graph gave her goosebumps.
“This is why I joined Eedi,” she said.
What do you anticipate your project accomplishing in 5 years?
We’re on a mission to help improve the learning outcomes of 1 billion children by 2030.
We believe that each one of those children will have their own personalized knowledge graph that is interoperable across platforms, based on an open-sourced unified knowledge graph created and maintained by Eedi.
Why stop there? There are 2 billion children in the world, many without classrooms in countries ravaged by war, and many more don’t own devices. We have a responsibility and a vision to help every child. Collaboratively, we can make a difference.
Five years? We would like to make a sizable step towards our goal for 2030!
Kent Fischer
Communications Director