Learning that adapts to you

Adaptive learning is a personalized learning approach where the path, pace, and content of a course change in real time based on each learner's performance. Instead of every student moving through the same lessons in the same order, an adaptive system assesses what each learner knows, where they're struggling, and what they need next — then routes them accordingly. Drivia's adaptive layer uses Bayesian Knowledge Tracing (BKT), Csikszentmihalyi flow-state theory, and a multi-model AI tutor (JAX) to do this without a human instructor having to manually re-sequence anything.

Personalized learning is the broader umbrella — any approach that tailors learning to the individual. Adaptive learning is a subset that does it algorithmically and in real time. A teacher hand-picking 5 different worksheets is personalized. A system that adjusts the difficulty of question N+1 based on how you answered question N is adaptive. Drivia does both — instructors can curate, and the platform adapts.

BKT is a probabilistic model that estimates the likelihood a student has mastered a specific skill based on their answer history. Instead of treating each quiz question independently, BKT updates a belief about each skill's mastery probability after every answer. Drivia uses per-quiz-question BKT to decide when to advance, when to remediate, and when to inject a spaced-repetition review. It's the same family of model that powers Khan Academy's mastery system and Carnegie Learning's MATHia.

Meta-analyses across 50+ studies (Steenbergen-Hu & Cooper 2014, Pane et al. 2017, Kulik & Fletcher 2016) show adaptive learning yields a 0.4–0.5 effect-size improvement over traditional one-size-fits-all instruction — roughly the difference between a C and a B. The gains are largest in math, science, and language-learning, and largest for learners in the middle of the achievement distribution. Drivia ships with these models pre-tuned for the course types it generates.

The H2E NEZ (Normalized Expert Zone) module classifies the learner's state into four bands every 60 seconds based on quiz performance, response latency, JAX interaction patterns, and lesson completion velocity. If you're in flow (Csikszentmihalyi's optimal challenge zone), JAX stays out of your way. If you're in anxiety, JAX scaffolds — easier questions, more examples, slower pace. If you're in boredom, JAX challenges — harder questions, deeper concepts, faster pace. This is the adaptive layer in action.