Prompt-Engineered Cognitive Apprenticeship for In-Service K–9 Teacher Learning

• In Event: AERA Roundtable Session Sunday 11:45 am - Gold 1
  In Roundtable Session: Innovations in Teacher Learning: Cognitive Apprenticeship, Feedback, and Technology Integration

Abstract

This study investigates the use of large language models (LLMs) in teacher professional development through a prompt-engineered framework in cognitive apprenticeship. Implemented in a 10-week asynchronous programming course for K–9 educators, the design simulates instructional phases—modeling, scaffolding, coaching, and reflection—via structured Custom GPT interactions. Design-based methods and qualitative analysis of hundreds of logs revealed (1) a 14% model error rate, (2) frequent student overreliance without constraints, (3) successful orchestration of instructional roles via prompt design, and (4) institutional feasibility and FERPA-compliant deployment. The framework preserves instructor authority as the primary evaluator, ensuring alignment with curricular goals. Findings contribute a tested blueprint for ethical, scalable, dialogic use of LLMs that retain instructor control while encouraging learner reflection.

Authors

• Alberto Cureg Cruz, California State University, Bakersfield

• Anjana Yatawara, California State University, Bakersfield

• Maruti Mishra, California State University, Bakersfield

• Jianjun Wang, California State University - Bakersfield