How adaptive testing chooses the next question
· 2 min read
A fixed test asks every student the same questions. An adaptive test asks each student the questions that are most useful for them.
What the research describes
Wainer and colleagues’ primer describes how to build and use a computerized adaptive test, which estimates a test taker’s level from the answers so far and chooses the next questions to suit it.
This work is about measurement, not about SubjectPath.
Why adaptivity helps a diagnostic
A question that is far too easy or far too hard for a student says little about them. Choosing questions near a student’s current level means each answer carries more information, so the diagnostic can spend its questions where they tell it the most.
How SubjectPath applies this idea
An Algebra Readiness session opens with a 15-question screener across the math domains it covers. Then come up to 25 targeted follow-up questions, focused on the domains where the diagnostic needs more information. After each answer, it updates its estimate and picks one of the questions expected to tell it the most. A session never runs past 40 questions, and it ends sooner once another question would add little precision.
Adaptive isn’t the same as easier
An adaptive test doesn’t make things easier for a student who is struggling or harder for a student who is doing well, as a reward or a penalty. It chooses questions that tell it the most about where each student is. For a student whose answers so far are mostly correct, that tends to mean harder questions; for a student who has missed several, easier ones.
Why a limit matters
An adaptive test could, in principle, keep asking questions for a long time. Algebra Readiness never asks more than 40 in a session, and it stops sooner when another question would add little precision.
Source
- Wainer, H. (Ed.). (2000). Computerized adaptive testing: A primer (2nd ed.). Lawrence Erlbaum.