Principal Investigator
Christina Weiland and Jason Corso – University of Michigan
Project Description
This study develops and tests a multi-modal artificial intelligence (AI) tool designed to analyze Pre-K classroom video and audio data to identify instructional practices that predict children’s learning outcomes. Using existing classroom observation data and child-level assessments from a large urban Pre-K program, the study will train and validate AI models to capture both traditional and novel measures of instructional and interactional quality. The research includes tool development, testing against established observational measures, and analysis of how AI-derived instructional practices relate to gains in children’s language, literacy, math, executive function, and social-emotional skills. The project aims to create scalable, cost-effective tools that provide more reliable and actionable measures of classroom quality and inform future coaching and professional learning systems.
Research Questions
- To what extent can AI systems generate reliable and valid measures of Pre-K instructional and interactional quality?
- How do AI-derived measures compare to traditional observational measures of classroom quality?
- Which instructional practices identified by the AI system are most strongly associated with improvements in children’s learning outcomes?
