Phase I CAMEL-CN: Modeling the Messy: Capturing Student Reasoning within High-Dimensional Industrial AI Datasets

There have been calls for more authentic and complex mathematical modeling experiences for high school students to better prepare students for STEM careers. In this project, a team of engineers, data scientists, science of learning researchers, and teachers collaborate to turn real-world, complex datasets into materials for the classroom and produce a large, AI-ready dataset on how students learn mathematical modeling. Data from high school students across multiple school districts are collected related to the student learning process and achievement in statistical reasoning, algebraic modeling, and data literacy. The project aims to provide a scalable, shareable framework for turning real-world, complex datasets into materials for the classroom, and to produce an AI-ready Student Math Modeling Learning Dataset that can be used for research and developing educational technologies. The project develops a new educational dataset that captures how students reason through complex, authentic, industrial data in open-ended computational modeling tasks. The work involves preparing industrial datasets on building cooling systems and advanced electronics for the classroom, creating infrastructure for collecting data on student learning, and building teachers’ capacity to use real-world datasets for classroom instruction. The investigators collect data from 4500 students across multiple school districts, yielding tens of millions of events. A key project goal is to use these data to distinguish productive academic struggle from unproductive frustration and convey this information through teacher dashboards. In addition to the learning process data, there are also pre/post assessments, student work, and classroom observation data. These data will be shared in FAIR-aligned repositories so that they can be used by researchers to better understand math learning and education. For curricular materials, a Data Translation Framework and protocol will be developed that consists of: 1) decision rubrics for data selection, 2) scaffolding templates, 3) learning analytics integration specifications, and 4) validation checklists. This will allow the development of new authentic datasets to use in mathematical modeling instruction. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria. NSF Award ID: 2621173 | Program: 4082PYXXDB NSF TRUST FUND | Principal Investigator: Rebecca Napolitano | Institution: Pennsylvania State Univ University Park, UNIVERSITY PARK, PA | Award Amount: $1,508,507 View on NSF Award Search: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2621173 View on Research.gov: https://www.research.gov/awardapi-service/v1/awards/2621173.html