Abstract: This study will investigate how variations in curriculum navigation and behavioral engagement relate to short-term learning gains, and to build machine learning models that predict immediate learning outcomes. We detail a two-phase approach: first, identifying and quantifying different learning pathway profiles among participants; second, training and evaluating predictive models to forecast post-assessment performance. This project will yield insights into which engagement behaviors and content navigation strategies correlate with better short-term learning in the Math Matrix program, and demonstrate how AI techniques can enhance educational outcomes by early identification of at-risk learners.
Principal Investigators:
Shuman Wang is a Ph.D. student in curriculum and teacher education at Stanford University.
Ari Jiayu An is a master’s student in education data science at Stanford University.
Abstract: This study employs Machine Learning (ML) techniques to predict math achievement trajectories for striving learners by analyzing engagement and persistence patterns within the i-Ready platform. Despite the potential of personalized instructional platforms to accelerate academic growth, significant variability exists in outcomes among users, some students make remarkable gains while others stagnate. By developing predictive models based on early-stage engagement indicators and persistence behaviors, this research distinguishes students likely to accelerate from those at risk of stagnation. The findings might provide educators with actionable insights for timely, targeted interventions and support Curriculum Associates in enhancing i-Ready’s responsiveness through improved early-warning tools. This work addresses both practical and methodological gaps in educational research, demonstrating how predictive analytics can transform support for striving learners when intervention matters most.
Principal Investigators:
Eter Mjavanadze is a doctoral student at George Mason University.
Angela Miller is an associate professor in research methods and educational psychology at George Mason University.
Abstract: This study investigates teacher engagement in the Math Matrix Digital Learning Platform (DLP) to understand factors that drive or hinder motivation. Situated expectancy-value theory (SEVT) is a motivational theory suggesting that an individual’s perceived competence and values predict decision-making and outcomes. Thus, we will use SEVT to examine teachers’ takeaways and challenges of using the DLP and how engagement and teacher characteristics (i.e., grade level, gender, race) predict competence beliefs, perceived usefulness, instructional strategy implementation, and ability to motivate their students. Findings will inform instructional design of the Math Matrix DLP to ensure effective teaching practices and student learning. This study will contribute to the broader conversation about the effectiveness of DLPS as research infrastructure. Leveraging platform-generated data will provide insights into the relation between teacher engagement with the platform and teacher motivation and whether patterns vary based on important teacher characteristics (e.g., grade level taught, experience level, etc.). These findings will not only inform the refinement of the Math Matrix DLP but also aligns with the goals of the DLP to offer guidance for the broader design and implementation of DLPs in mathematics education. The findings from this research will offer valuable insights into how DLPs can be optimized to improve mathematics instruction through enhancing teacher motivation, ultimately supporting student learning outcomes.
Principal Investigators:
Dr. Patrick Beymer is an assistant professor of psychology at the University of Cincinnati.
Dr. Jessica Gladstone is an assistant professor of educational psychology at the University of Illinois, Urbana-Champaign.
Abstract: The project aims to develop methodology and algorithms that optimize learning paths for improved efficiency. It will integrate robust measurement and knowledge tracing models to estimate students’ abilities along with their associated measurement errors. A Deep Q-Network-based recommendation system will guide learning by balancing skill improvement with reductions in measurement uncertainty, resulting in highly personalized learning paths. In addition, the proposed methodology will help us understand transfer of learning to new topics and estimate rates of acquisition and forgetting over time. This work directly supports PI’s (Yikai EK Lu’s) trajectory as an early-career researcher transitioning into a tenure-track assistant professor role.
Principal Investigators:
Yikai “EK” Lu is a doctoral candidate at the University of Notre Dame.
Dr. Ying (“Alison”) Cheng is professor of psychology at University of Notre Dame.
Abstract: Knowledge tracing (KT) predicts learning performance by analyzing past behaviors to enable adaptive instruction and precision education, particularly within digital platforms such as Math Matrix Micro-Credential. Traditional KT models primarily rely on sequential patterns (e.g., temporal sequence of assessment responses), often downplaying valuable semantic insights from textual data, such as problem similarity, embedded concepts, and prerequisite relationships. Meanwhile, prior KT methods have largely prioritized predictive accuracy, offering limited transparency into the underlying decision-making processes and leaving educators with few actionable insights. Leveraging the rich assessment data, open-ended responses, and user behaviors from Math Matrix Micro-Credential, we propose TRACELLMs: an explainable KT framework that fuses sequential interaction data, assessment text, and teachers’ historical learning records to evaluate conceptual understanding and deliver actionable feedback. We will benchmark TRACE-LLMs against existing KT methods to demonstrate its feasibility, robustness, and scalability. Successful validation of TRACE-LLMs will enable its integration into Math Matrix Micro-Credential for precise tracking of teacher competency growth, AI-powered mentoring, and personalized learning pathways, ultimately advancing high-quality, adaptive, and scalable professional learning experiences.
Principal Investigator:
Dr. Chenglu Li is an assistant professor of learning sciences at the University of Utah
Abstract: Carnegie Mellon University researchers are developing a new approach to enhance student engagement and persistence in mathematics using the i-Ready adaptive learning platform. Led by PhD student Conrad Borchers in collaboration with Co-PIs Vincent Aleven, Ken Koedinger, and Danielle Thomas, the project explores how regular, personalized goals and feedback can help middle school students stay motivated and on track toward long-term proficiency growth targets. Building on prior success from the PLUS tutoring project, where goal setting increased engagement by 25% and skill mastery by 40%, this initiative seeks to design scalable goal-setting tools integrated into i-Ready’s existing features.
The planning grant will support the creation of persistence analytics and adaptive goal recommendations using i-Ready’s lesson and diagnostic data. In partnership with Curriculum Associates and a school district to be selected, the project aims to empower teachers to guide students in setting and achieving goals, including through i-Ready’s “Data Chats” and growth tracking tools. Ultimately, the project advances research on student motivation and self-regulated learning while aligning with AIMS EduData’s mission to generate actionable insights from digital learning platforms.
Principal Investigator:
Conrad Borchers is a Ph.D. student at the Human-Computer Interaction Institute (HCII) at Carnegie Mellon University’s School of Computer Science
Abstract: KIPP Team and Family, in partnership with KIPP New Jersey (Camden), will conduct a research-practice partnership with Khan Academy to evaluate AI-powered math interventions for underperforming students in high-variance classrooms. The partnership will investigate how Khanmigo can be leveraged alongside I-Ready diagnostic data to create targeted interventions that address this extreme variance, particularly for students performing below grade level within the Illustrative Mathematics curriculum, among other research questions.
Principal Investigator:
Kevin Shaw is the director of AI, innovation, and strategy at KIPP Team and Family
Abstract: This project explores how teachers interact with and benefit from online professional learning (PL), particularly in mathematics education. As more educators enter teaching through various pathways, ongoing PL becomes essential for ensuring high-quality instruction. Our project focuses on understanding teachers’ engagement, how they participate, think about, and value their learning experiences, and its impact on their teaching effectiveness. We’ll use survey and online activity data to identify patterns of engagement and connect them to teachers’ learning outcomes and perceptions. Insights from this project will help create better, more impactful online PL experiences that enhance mathematics teaching and student success.
Principal Investigators:
Dr. John Chukwunonso Ojeogwu is a STEM education postdoctoral researcher at the Desert Research Institute (DRI) and the University of Nevada Las Vegas.
Erin Smith is an associate professor of mathematics education at University of Nevada, Las Vegas.
