Abstract
This three‑manuscript dissertation investigates how artificial intelligence (AI)–informed, video‑based classroom activity data can support teacher noticing across multiple stages of the teaching profession. Grounded in research on teacher noticing and the design of learning analytics tools, the dissertation examines how elementary teachers, pre‑service teachers, and instructional coaches interpret and use automated dashboards to recognize, reason about, and respond to instructional activities.
Manuscript 1 presents a qualitative case study of six in‑service elementary teachers who engaged with the Artificial Intelligence for Advancing Instruction (AIAI) dashboard, a tool that uses neural network models trained on 244 hours of classroom video to classify instructional activity structures. Findings show that teachers used the dashboard to attend to and elaborate on their instructional decisions, compare activity structures across lessons, and reflect on patterns not easily visible during live instruction. Teachers valued the dashboard’s non‑evaluative visualizations, expressed a desire for more frequent access to automated data than traditional observations allow, and preferred sharing analytics rather than raw video with colleagues.
Manuscript 2 extends this work to pre‑service teachers by examining how seven teaching candidates perceived and used the Classification of Reading and Math Instruction (CERMI) dashboard. Using deductive and inductive analysis of interviews and dashboard reflections, the study identifies four themes: noticing and reflection, willingness to share data, timeliness of feedback, and prior experience with technology. Candidates reported that the dashboard made instructional patterns more visible, supported evidence‑based interpretation of classroom events, and increased their awareness of student engagement and activity structures. The findings highlight design features—non‑evaluative analytics, straightforward visual displays, and video‑linked metrics—that help novices develop professional vision.
Manuscript 3 investigates how four instructional coaches and their four teaching candidates used the CERMI dashboard to support video‑based coaching. Drawing on the Learning to Notice framework, the study analyzes how participants used lesson summaries, comparative pie charts, and reflection prompts to recognize significant events, reason with contextual knowledge, and connect analytics to broader principles of teaching and learning. Coaches used the dashboard to scaffold noticing, calibrate interpretations, and support goal setting, while candidates used it to identify instructional patterns and consider alternative pedagogical moves. Variation in perceived accuracy and prior experience with AI shaped participants’ trust and uptake of the tool.
Across the three manuscripts, this dissertation demonstrates that AI‑informed dashboards can meaningfully support teacher noticing when dashboards are designed to be non‑evaluative, interpretable, and tightly connected to video evidence. The findings contribute to emerging scholarship on AI‑supported teacher learning and offer design principles for developing scalable, human‑centered analytics tools that enhance reflection, coaching, and instructional decision‑making in teacher education and professional development.
