May 2019
Following the annual edtech investor conference at ASUGSV, it's been an exciting time in the edtech industry, with notable deals and investments for Coursera, Labster, Chegg, Gaosi, Trilogy and FutureLearn, among others.
At Adaptemy, we have made a major breakthrough in our learner model using some of the latest AI technologies. With a model accuracy of 0.913 (AUC metric), we are clearly setting a new benchmark for the industry. For our partners, this more accurate model gives product designers new capabilities to create better personalised learning products.
How?
Our major increase in accuracy is based on innovations in 3 parts of our model:
- A self-learning curriculum map, based on maps created in our Curriculum Design tool, that uses machine learning techniques to infer learning patterns. This is some of the most advanced AI in education.
- A content model that calculates metrics for the quality, difficulty, effort and engagement profile of each piece of content, and updates continually based on learner data.
- A learner data gathering and cleaning stage that more accurately accounts for demotivated and unfocused activity that does not reflect the abilities of the learner.
These 3 elements are now integrated into a continuous feedback loop, improving the system accuracy with each new learner interaction.
Our research will be presented at various conferences this year, including CSEDU this week and UMAP (Conference on User Modelling, Adaptation and Personalisation) in June.
We look forward to building the next generation of effective learning products with our partners. See some examples of what we have built, or get in touch to set up a demo with your team.
Conor O'Sullivan
CEO at Adaptemy
|