2026 ASEE Annual Conference & Exposition

Optimizing Feedback Loops in Adaptive Learning Systems: A Computational Model-Based Integration of Diagnostic Analytics and Cognitive Load Principles

Presented at DSAI-Session 1: Data Science Program Design and Curriculum Frameworks

Adaptive learning in engineering education has long excelled at error detection, yet research on the systematic transformation of diagnostic insights into pedagogical actions remains rare. This study advances a design-based, mixed-method inquiry grounded in systems-engineering epistemology and informed by model-based reasoning and cognitive load theory. The proposed diagnostic–pedagogical framework unifies Item Response Theory (IRT), Cognitive Diagnostic Modeling (DINA), and Large Language Models (LLMs) within a coherent architecture of adaptive instructional design. IRT estimates learner readiness along continuous ability trajectories, DINA models the structural interdependence of skills, and LLMs contextualize diagnostic evidence by mapping conceptual relationships across course-specific multimedia resources. Through greedy and gradient-based orchestration, the framework operationalizes adaptivity as an optimization of learning flow, allocating resources that balance diagnostic precision with cognitive economy. The system generates evidence-based feedback grounded in an awareness of cognitive fatigue that draws on prior learning materials and encourages conceptual refinement rather than repetitive review. This model was tested using 1,000 simulated and 726 actual Engineering learners, and validated through student interviews and expert reviews, demonstrating that adaptive learning can function such that psychometric rigor, instructional design, and human judgment coalesce to sustain meaningful, cognitively attuned progress.

Authors

Mr. Amirreza Mehrabi Purdue Engineering Education [biography]

Dr. Amirreza Mehrabi is an accomplished scholar at Purdue University specializing in machine learning and AI-aided modeling and Engineering Education. As an interdisciplinary researcher, he designs and deploys intelligent platforms that tackle complex decision-making challenges, integrating advanced techniques such as transformer models, reinforcement learning, and statistical inference. His work spans across adaptive testing, cognitive fatigue detection by ML, agentic AI agents tutors, and multimodal analytics, leading to innovative solutions with measurable real-world impact. Amirreza’s expertise is reflected in his custom architectures, model optimization research, and psychometrics—combining educational measurement science with deep learning. Amirreza’s collaborative spirit, technical mastery, and commitment to transparent, bias-aware modeling foster a vibrant culture of innovation.
Dr. Jason W Morphew http://orcid.org/0000-0001-5971-214X Purdue University – West Lafayette [biography]

Dr. Jason Morphew is an assistant professor at Purdue University in Engineering Education and serves as the director of undergraduate curriculum and advanced learning technologies for SCALE, and is affiliated with the INSPIRE research institute for Pre-College Engineering and the Center for Advancing the Teaching and Learning of STEM. Dr. Morphew's research focuses on the application of principles of learning derived from cognitive science and the learning sciences to the design and evaluation of learning environments and technologies that enhance learning, interest, and engagement in STEM. Morphew earned his Ph.D. from the University of Illinois Urbana-Champaign in Educational Psychology from the Cognitive Science of Teaching and Learning Division where he examined self-regulated learning of engineering students enrolled in introductory physics courses.
Breejha Sene Quezada Purdue University
Prof. N. Sanjay Rebello [biography]

Dr. Sanjay Rebello is Professor of Physics and Astronomy and Professor of Curriculum and Instruction at Purdue University. His research focuses on the teaching and learning of physics. He is particularly interested in issues pertaining to transfer of le

Note

The full paper will be available to logged in and registered conference attendees once the conference starts on June 21, 2026, and to all visitors after the conference ends on June 24, 2026

« View session