Poster #134 - Same Law, Different Systems: How IEP Data Infrastructure Shapes the Implementation and Measurement of IDEA

• In Event: Saturday Poster Luncheon

Abstract

The Individuals with Disabilities Education Act (IDEA) establishes a federal framework to ensure that students with disabilities receive individualized and appropriate educational services. IDEA standardizes required components of Individualized Education Programs (IEPs), which include present levels of performance, student needs, annual goals, and services. However, the software systems used to document and manage IEPs are decentralized and vary widely across states and districts. This study examines how variation in IEP data infrastructure, particularly software platforms and templates, shapes the representation and measurability of IDEA implementation.

We ask: To what extent does variation in IEP data systems affect the ability to measure substantive compliance with IDEA across districts? Substantive compliance requires that IEP components are logically connected and aligned to student needs, yet evaluating this alignment at scale is challenging when data are structured differently across contexts.

To address this challenge, we apply SEAMLESS (Special Education Applications of Machine Learning to Enhance Student Success), an AI-enabled approach that extracts and standardizes core IEP constructs from unstructured text. Using a dataset of approximately 2,000 IEPs from four districts across two states, we analyze nearly 26,000 narrative passages to investigate the extent to which IDEA-required constructs are encoded differently across software platforms. We classify IEP content into a common schema (e.g., present levels, needs, goals, services) using natural language processing and machine learning to reconstruct the relationships among these components independent of document structure. This approach enables a proof-of-concept analysis of substantive compliance, including the extent to which student needs are supported by present levels and aligned with goals and services.

To date, preliminary findings suggest that differences in IEP structure significantly hinder cross-district comparability and may obscure meaningful variation in special education practice and outcomes. We have identified substantial complexity in how key constructs are segmented, labeled, and ordered within IEP documents. For example, in some districts, student needs are explicitly defined; in others, they are embedded within present levels or distributed across multiple sections. These differences do not necessarily reflect variation in policy or practice, but rather variation in how software systems structure and store information. By contrast, construct-level standardization enables consistent measurement of alignment across heterogeneous systems. These findings highlight a critical but underexamined source of policy variation: data infrastructure. Even under a uniform federal law, differences in state and district systems can shape what is measurable and, therefore, what is governable. Improving the comparability of IEP data at scale, either through standardization or translation tools, has important implications for policy evaluation, equity monitoring, and data-informed decision-making in special education.

Author

• Adrienne Danielle Woods, SRI International
  Presenting Author