What Is Adaptive Learning? How Personalized Learning Paths Work

Walk into almost any classroom and you will find thirty learners at thirty different points. Some mastered the topic last week and are quietly bored. Others never quite grasped the step before it and are now lost, and the teacher often finds out only when the end-of-unit test comes back. Teachers differentiate as best they can, but preparing several versions of every lesson by hand does not scale.

Adaptive learning tackles that problem with software that behaves a little like a good tutor working with one student: it watches each answer, decides what to try next and knows when to step back to an earlier idea. The difference is that it can do this for every learner in the room at once.

What Is Adaptive Learning?

Adaptive learning is an approach to teaching in which a digital platform adjusts the content, difficulty, pace and order of learning for each learner, using evidence from their own answers. Instead of everyone moving through the same sequence on the same dates, each learner gets the next step that fits what they have just shown they know.

The idea is older than the software. In the 1960s the educational psychologist Benjamin Bloom described mastery learning: teach a unit, check understanding, give corrective work to those who need it and move on when the skill is secure. Intelligent tutoring systems and computerized adaptive tests carried related ideas into software decades ago. What has changed is that a modern platform can apply it continuously, for every learner, across a whole course.

It helps to be clear about what adaptive learning is not:

• Not just self-paced. Letting learners click through at their own speed changes the pace but not the path. Adaptive learning changes what comes next.
• Not a single branch. A quiz that sends every wrong answer to the same remedial page is one fixed rule. An adaptive engine keeps updating its picture of the learner with every response.
• Not a replacement for teachers. It takes on routine sequencing and marking, so teachers can spend their time where judgment matters.

How an Adaptive Learning Engine Decides What Comes Next

Most adaptive platforms run the same loop. The learner attempts something, the engine updates its estimate of what the learner knows, and that estimate decides the next activity. Four mechanisms do most of the work.

Mastery Estimates for Each Skill

The engine does not hold a single grade. It holds an estimate for each skill in the course, such as finding equivalent fractions or identifying the subject of a sentence. A widely used family of methods, called knowledge tracing, goes back to work by Corbett and Anderson in the mid-1990s: the system keeps a probability that the learner has mastered each skill and updates it after every answer. Good models allow for guesses and slips, so one lucky answer does not prove mastery and one careless mistake does not erase it.

Diagnostics and Placement

Before the engine can adapt, it needs a starting point. A short placement assessment samples the skills in a unit and seeds the learner's profile, so nobody repeats topics they already know. Good diagnostics are adaptive too, moving on as soon as the evidence for a skill is clear.

Difficulty and Sequence

With an estimate for each skill, the engine chooses what comes next by asking two questions. Is the learner ready for this skill, because its prerequisites are secure? And is this item at the right level of challenge? Psychologists call the band of tasks a learner cannot yet manage alone but can manage with some support the zone of proximal development. In practice it is where effort pays off and learners stay engaged.

When a prerequisite is weak, the engine runs a remediation loop: it steps back to the earlier skill, often with a worked example or a hint, and then returns to the main path. When a learner is clearly ahead, it offers stretch content rather than more of the same.

Spaced Practice and Review

Learning something once is not the same as keeping it. A major review of study techniques by Dunlosky and colleagues gave its highest rating to just two of the ten techniques it examined, practice testing and distributed practice, because they helped learners of different ages and abilities across many kinds of test, including in real classrooms. An adaptive engine applies both: it schedules short retrieval checks on skills at risk of fading and spaces them out over time, instead of leaving revision to the week before an exam.

A Worked Example: One Learner, One Unit on Fractions

Here is how the loop plays out for an upper primary learner starting a unit on adding fractions. The details vary by platform, but the shape is typical.

1. Placement. A short diagnostic shows that the learner can simplify fractions but is unsure about equivalent fractions, a prerequisite for adding fractions with different denominators.
2. Remediation loop. Instead of starting the new topic, the engine offers a short lesson on equivalent fractions with a worked example, then a few practice items of rising difficulty.
3. Hints, not answers. When the learner hesitates, a hint points to the step they missed. Some engines give less credit to an answer reached after a hint, so the mastery estimate stays honest.
4. Back to the main path. Once the estimate for equivalent fractions is secure, the engine returns to adding fractions, starting with like denominators and then moving to unlike ones.
5. Teacher view. On the dashboard, the teacher sees that this learner and several classmates needed the remediation loop, and opens the next lesson with a short recap for that group.
6. Spaced review. Over the following weeks, short review items on equivalent fractions return at widening intervals, so the skill stays secure after the unit ends.

Nothing in that sequence is magic. Each step is something a good teacher would do for one learner, and the software makes it possible for a whole class at once.