ATLAS research fellows advance innovative assessment approaches with AI and adaptive testing
LAWRENCE — Assessments that adapt to each student’s abilities in real time. Artificial intelligence that makes assessments more accessible for English language learners with significant cognitive disabilities. Research fellows at Accessible Teaching, Learning, and Assessment Systems (ATLAS) are exploring innovative ways to make assessments more accessible and efficient.
Every year, ATLAS, a center within the Achievement & Assessment Institute at the University of Kansas, sets research priorities for its fellowship program. This year’s priorities were AI applications used within an operational assessment and the evaluation of longitudinal diagnostic classification models.
“The fellowship lets us expand the scope of what's getting done to help us move toward our mission,” said Jeff Hoover, an ATLAS psychometrician. “New people bring new perspectives and experiences, so creative and innovative solutions can get introduced.”
This year’s ATLAS research fellows are Victoria Quirk and Pragati Maheshwary. Quirk is a fourth-year doctoral student studying educational measurement and quantitative methods at the University of Illinois Urbana-Champaign. Maheshwary is a first-year doctoral student at the University of Wisconsin–Madison studying educational psychology.
Quirk will investigate hierarchical models within computerized adaptive assessments. Adaptive assessments use algorithms to personalize the questions to each student’s ability as they test, which can shorten testing times and provide better data for educators and researchers.
Hierarchical models determine what skills the student needs to accurately answer assessment questions. In loose hierarchical models, students do not necessarily have to know how to do one thing to do the next, while in strict models, they do. Quirk will compare these two models and study their impact on student outcomes.
“This is a really exciting area of study,” Hoover said, “because adaptive tests help get a very fine-grained takeaway on the skills students know well and areas where additional learning could be beneficial.”
As part of the AI research priority, Maheshwary will research the use of human-informed AI systems for early-stage item drafting including the development of assessment items in multiple languages for English learners with significant cognitive disabilities.
The process of item generation, or question writing, is time consuming for test developers. Maheshwary will look at ways to streamline the process while keeping humans in the loop and meeting alternate assessment accessibility standards. The inclusion of AI would also improve accessibility through the generation of Spanish, Spanglish and simplified English variants.
“The proposal is a very interesting and thoughtful way to start thinking about AI as a tool and how to use it well,” Hoover said. “This research gets the ball rolling on incorporating those tools into our workflow in a way that's efficient and still produces high-quality products.”
Quirk and Maheshwary will work with ATLAS research staff, who will help guide their research and provide feedback. Both fellows will complete their work in spring 2026.
“I'm honored to have been selected,” Quirk said. “I hope that the work will be something that is both instructionally and operationally useful, not only for educators and students, but also useful for test creators.”
Maheshwary is especially eager to use AIBAT, a tool developed within the TRAIL lab at UW-Madison, to enhance her work with ATLAS.
“I am excited to bring in tools from two different universities to test them and hopefully expand on them,” Maheshwary said. “This is a great opportunity to mix cutting-edge technology that incorporates AI and bring it into the assessment field.”