Definition · Adaptive learning
Students & teachers Educational researchers & ed-tech developers

What Is Adaptive Learning? A Precise Definition.

By Gregory Stuart Lacefield · May 2026 · gregorylacefield.com

The precise definition

Adaptive learning, precisely defined, is an instructional system that continuously updates its model of a specific student's understanding and uses that model to select what to teach next, at what difficulty level, and in what context. The key word is continuously — not once at intake, not occasionally based on test scores, but after every interaction. Most platforms marketed as adaptive are not adaptive in this sense.

A system that asks a student twenty questions and routes them to one of three content tracks is not adaptive. It is branched. Adaptive means the model updates every time the student does something, and the next thing the student is asked reflects that update.

True adaptive learning requires three components operating simultaneously: a model of the student's current understanding (what they know, what they think they know but don't, and what they have never encountered), a structured map of what the subject area consists of and how concepts depend on each other, and a mechanism for matching the student's current state to the most productive next learning experience.

Most "adaptive" platforms have a simplified version of the third component and almost none of the first two.

What most platforms actually do

When an ed-tech company calls their platform adaptive, they usually mean one of three things: the platform adjusts pacing (letting students move faster or slower through fixed content), the platform routes students to easier or harder versions of the same content based on quiz scores, or the platform uses spaced repetition to schedule when items are reviewed.

These are useful features. None of them are adaptive in the full sense. They treat learning as a matter of exposure and recall — if the student sees the right content at the right time, learning happens. The deeper problem — that the student may have a specific misconception that makes correct understanding impossible regardless of how many times they encounter the content — is not addressed.

What genuine adaptation requires

A genuinely adaptive system must be able to distinguish between three different explanations for a student's wrong answer: the student has never been exposed to the concept, the student was exposed but didn't learn it, and the student learned something incorrect that is now interfering with learning the right thing. These require completely different responses. Giving the student more of the same content helps in the first case, partially helps in the second, and makes the third case worse.

This is the problem the Lacefield Adaptive Learning System is designed to solve. The diagnostic intake doesn't just measure what a student gets right and wrong — it probes for the specific misconception behind each wrong answer. A student who gets 3 + 4 = □ wrong has a different problem than a student who gets □ + 4 = 7 wrong, even though both problems involve the same arithmetic. The first student may have an arithmetic gap. The second student probably has an equals-sign schema problem — they read the equals sign as "the answer goes here" and are confused when it appears before the missing value.

The misconception behind the error is more important than the error itself. Correcting the error without addressing the misconception produces a student who gets that specific problem type right and fails on every variation.

Bloom's 2-sigma problem

In 1984, Benjamin Bloom documented that one-on-one tutoring produces results approximately two standard deviations better than conventional classroom instruction. The average tutored student outperforms approximately 98% of students receiving conventional instruction. This finding — known as the 2-sigma problem — has been replicated and extended extensively.

The question Bloom posed was whether these results could be achieved at scale. The answer, forty years later, is that no system has reliably achieved it. Adaptive learning platforms represent the most serious attempt — but most of them are not adaptive enough to replicate what a skilled human tutor actually does, which is maintain a continuously updated model of the specific student's understanding and make every interaction reflect that model.

That is the goal of the Lacefield Adaptive Learning System. Not to replace human tutors — in MVP 1, human tutors are the delivery mechanism — but to give human tutors the diagnostic framework, the concept map, and the data infrastructure to operate at the level of precision that produces Bloom's results.

Frequently asked questions

What is the difference between adaptive learning and personalized learning?
Personalized learning typically means adjusting the pace or format of instruction to the student's preferences — self-paced, visual vs. text, etc. Adaptive learning means adjusting the content, difficulty, and sequence of instruction based on the student's demonstrated understanding. Personalization is a surface adjustment; adaptation is a structural one.
What is a schema floor?
Schema floor is the highest level at which a student's understanding is genuinely sound enough to build on. A student can perform at an apparent higher level through pattern-matching while having a schema floor at a lower level — the higher-tier performance is unstable and collapses under novel conditions. Identifying the true schema floor is the primary function of the intake diagnostic in the Lacefield system. See the full definition in the technical glossary.
What is wrong schema?
Wrong schema is an apparent higher-tier understanding built on a corrupted lower-tier concept. The student has a model for the concept, but the model is incorrect in a specific identifiable way. Wrong schema is more dangerous than absent schema because it feels internally consistent to the student and resists correction. The most common wrong schema in algebra: the equals sign read as "the answer goes here" rather than as an assertion of equivalence.
What is cognitive load theory and how does it relate to adaptive learning?
Cognitive load theory, developed by John Sweller (1988), holds that working memory has a limited capacity. When that capacity is exceeded — by too much new information, or by basic operations that should be automatic but aren't — learning degrades or stops. Adaptive learning systems that calibrate difficulty correctly keep cognitive load in the productive range: enough challenge to require engagement, not so much that working memory is overwhelmed. The Lacefield framework addresses cognitive load directly through the foundational fluency principle: basic arithmetic must become automatic before higher reasoning is reliable, precisely because slow arithmetic consumes working memory needed for the concepts being taught.

Author: Gregory Stuart Lacefield — independent systems engineer, researcher, and educator. Creator of the Lacefield Adaptive Learning System.

Full author record · White papers & research · Technical glossary · Engineering specification