Adaptive learning

Adaptive learning is software that chooses each learner's next question from the answer to the last one. A learner who gets a question wrong gets an easier one, or more practice on the same idea. A learner who gets it right moves on to a harder one. So each person practises at the edge of what they know, instead of everyone doing the same worksheet in the same order. When the same method is used to measure a learner rather than to teach, it is called an adaptive test.

Here is how it works in general. Every question in the software's bank is tagged with the topic it tests and with how difficult it is. The learner answers a question. The software updates its estimate of how well the learner knows that topic, and the topics it depends on. For example, a child who cannot add fractions may first need practice on finding a common denominator. The software then picks the next question: one that the learner is likely, but not certain, to answer correctly. A question that is far too easy teaches nothing, and one that is far too hard only discourages. This repeats after every answer. A teacher sees a report of what each learner has and has not mastered.

An adaptive test works the same way, but its aim is a score. Each question is chosen to tell the software as much as possible about the candidate's level. So an adaptive test can reach a reliable score with fewer questions than a fixed paper.

Take, for example, a class 6 student using an adaptive learning platform for maths. After three wrong answers on decimals, the software goes back to place value, and returns to decimals once the student gets those questions right.

Consider a candidate taking an online English proficiency test at home. Each correct answer brings a harder question, so the test finds the candidate's level in about an hour.

For instance, a language learning app brings back the words that a user keeps getting wrong, and shows less often the ones the user already knows.

How is adaptive learning different from a mock test? In a mock test, every student gets the same paper, so that the scores can be ranked against each other. In adaptive learning, every student gets a different set of questions, so that each one practises what they need. Adaptive learning is also different from an AI tutor. Adaptive software picks from a bank of questions written in advance, and it does not explain in conversation. An AI tutor answers questions in its own words. Many products, such as digital courseware, now combine both.