From dialogue to differentiation: AI-agent-supported coaching for teacher efficacy with multilingual language learners

Abstract

Introduction The purpose of this study was to strengthen 5th grade elementary reading, math, and ESOL teachers' instructional expertise in integrating oral and written language development within Tier 1 differentiated instruction for multilingual language learners (MLLs), through a job-embedded professional development model for data-based decision-making (DBDM). Methods This mixed methods study explored outcomes across four dimensions: 1) coaches' perceptions of this model to improve teacher instructional efficacy with MLLs; 2) teachers' perception of the utility of an Artificial Intelligence (AI) agent to design curriculum-based rubrics; 3) teachers' efficacy with DBDM for MLLs; and 4) the implications of this process for improving English language proficiency of 5th grade MLLs. Coaches supported grade level teachers to differentiate Tier 1 curriculum in whole and small group settings. Through sustained professional learning, coaching, and PLC-based collaboration, 5th grade teachers learned to develop explicit language objectives with grade-level ELA, reading and math curriculum. An AI agent supported teachers by generating WIDA standards-aligned rubrics which teachers used to monitor real-time English language proficiency progress, inform differentiated instruction, and support decision-making processes for intensifying interventions. Results Outcomes of the study showed meaningful improvement across all four dimensions. Coaches reported growth in teacher efficacy planning and delivering instruction for MLLs, increases in data-driven conversations, and specificity in writing language objectives. Overall, teacher responses on quantitative surveys also corroborated these findings and rated the AI-generated rubrics as easier to use and understand. Finally, 5th grade MLLs exceeded their growth thresholds at a higher rate than previous cohorts. Discussion Results show promise for: 1) employing the Caminos Data Dialogue Protocol for Multilingual Learners as a format for job-embedded professional development; 2) AI agents to function as mediated scaffolding, bridging coaching to teacher independence; and combined, 3) this model was associated with strong English language proficiency gains for 5th grade MLLs, indicating the importance of differentiating Tier 1 instruction. By strengthening both educator capacity and access to language data, this study contributed to the development of sustainable, school-embedded systems of support for MLLs with diverse language and learning profiles.