After more than two years of ‘pandemic life’, digitalisation has conquered our daily lives. Everything is done through the internet, from shopping to meetings to learning – and testing has not been spared. Andrew Stokes wrote an article showing how digitally-based tests are more preferable than paper-based tests. The question is, what more can be done for tests other than just putting the test online with some fancy layouts?
Adaptive testing is an innovative and results-driven feature that takes online testing into a new era. Adaptive tests not only give a refined result to the administrator but also present a more personalised test difficulty for the test taker. These tests automatically and dynamically generate test plots – a feature we should aim to have in all our tests. In this post, I want to explain clearly and simply how adaptive testing works and what it means for the future of testing, in the context of the Dynamic Placement Test.
How does adaptive testing work (technically)?
There are two key features to adaptive testing. The first is difficulty labels. All questions in our question bank are marked with a difficulty level. The question tests takers see depends on whether they pass or fail that difficulty level.
The second feature is checkpoints. In the Dynamic Placement Test, we have implemented multiple checkpoints in between questions where the test decides whether test takers need to move up a level, move down, or remain at the same level. This is all done in real time, making each test personalised to the test taker, adapting to their abilities.
When does the test know to give students a harder question, easier question or the same level question?
The test begins with a set of 10 random questions. When the test taker submits their answers, the first checkpoint does the marking immediately, then determines what difficulty the next set of questions will be. For example, if the test taker gets only 10% of the questions correct, the test will lower the difficulty of the next questions. If the test taker gets 50% of the questions correct, the test will maintain the difficulty level. If the test taker gets 100% of the questions correct, the test will introduce more challenging and higher level questions.
The system then draws questions, at random, from the question bank of that difficulty. This pattern and logic continues until the test taker reaches the end of the test where their CEFR level is determined.
You can read about the Dynamic Placement Test’s test construct in more detail here.
Why is adaptivity good for the test? Are more technically advanced tests better and more interesting?
An adaptive test can provide a more accurate and refined result without the test-taker doing a huge number of questions. It dynamically selects a similar level of questions according to the ability of the test taker, without having questions that are too simple or too hard. The result? The test taker’s specific level within a reasonable boundary of time. In contrast, a traditional paper test requires the test taker to complete a large amount of questions, including some questions that will be irrelevant to their level.
You can read more about the validity of the Dynamic Placement Test here.
