Forum Discussion
Adaptive Reading Diagnostic
I think I finally figured out how to create an adaptive skill assessment! I would love any feedback you have to take it to the next level.
This assessment is designed to assess reading level, and to do so using adaptive testing that dynamically responds to the user’s responses in about 10 questions.
In other words, if you answer a question correctly, the program then asks you to answer a more difficult question. If you answer a question incorrectly, the program asks you to answer a less difficult question. As the program learns where the peak of your ability is, it determines your current skill level.
The adaptive branching logic for this diagnostic is handled through JavaScript so that the user’s experience is simplified to a single interface. I enlisted AI to help with the JavaScript--it really simplified the process.
The program is designed to look for multiple correct answers before it determines a skill level. That means no single right or wrong answer determines the score—it can handle outliers and establish a larger pattern of consistent ability. That also means the program incorporates a question bank that is large enough so that the user does not see repeat questions as they move through.
If you're interested in looking under the hood, let me know and I'll attach the storyline itself!
3 Replies
Very cool jegagnon!
I had a little bit of a standardized testing flashback and worked my brain in a way it hadn't been used in a minute. 🤭
I think it would be great if you are willing to share the Storyline file. Also, how did you decide on 10 questions as the sweet spot?
SlimMasmoudi this might be interesting to you, given your question a few months back!
The storyline file is here (I hope).
10 questions is the average user experience: most players use about that many questions.
I decided on 3 correct answers as enough of a streak to have some confidence in the result, and because there are 5 skill levels, the question bank actually contains 40 questions (8 at each level). Statistically, more than 3 would be better, but that also lengthens the diagnostic, so I settled on 3 as a reasonable compromise (also I was using this as an initial diagnostic and not a summative assessment).
3 correct also means that the program tested the player with 2 harder questions and the player got both wrong.
Katie-Jordan, thanks
