- WHITEPAPER -
Prescribing Deep Attentive Score Prediction Attracts Improved Student Engagement
"The first forward-looking research to prove the real world value and commercial effects of deep learning prediction models with data"
Accepted as a full paper at Educational Data Mining Conference (EDM) 2020
A case study verifying that deep learning AI creates user value and drives tangible business growth in real life, not in the lab.
In one study, we conducted an A/B test by randomly administering two different score prediction algorithms to 78,000 Santa users: one based on a collaborative-filtering algorithm and another based on deep-learning. The results showed that, based on Mean Absolute Error, the deep-learning model induces greater student engagement measured by metrics such as the average number of questions a student answered after the diagnostic test and the percentage of students who converted to paid subscriptions.
The data demonstrate that accurate score predictions based on deep learning have a positive impact on user confidence in learning tools, increasing their learning motivation.