Abstract:

AI Partnered Supervision: A Multi-Stage Generative Workflow:
Supervision of teacher residents is labor-intensive, with traditional methods causing workload bottlenecks and limiting deep analysis. This project introduces the **Islander Growth Lens (IGL)**, a Standards-Aligned Growth and Feedback AI-Driven Platform featuring an advanced, multi-stage generative workflow. IGL summarizes full-semester observations, extracts cross-case patterns, and analyzes supervisor feedback tendencies to enhance reflective practice.

Phase 1: Automated Documentation:
The initial phase reduced the time supervisors spent drafting each observation from 20–30 minutes to **8–12 minutes** (a measurable **40–60% reduction**). IGL automatically generates structured, polished drafts aligned to program domains (R+, Reinforcement; /R-, Refinement evidence, UDL, actionable feedback) using consistent coaching language, directly addressing documentation workload.

Phase 2: Full-Semester Pattern Detection:
Addressing the difficulty of tracking longitudinal growth across dozens of notes, IGL employs a staged summary approach to process large collections of data while preserving detail and chronology. This summary then enables pattern-detection analysis. The system identifies resident strengths/challenges across key instructional domains (e.g., planning, questioning) and extracts student-perception indicators (e.g., confusion, enthusiasm) from notes and conference transcripts. This synthesis generates case-specific insights and cross-resident themes often missed under time pressure.

Phase 3: Supervisor Self-Awareness:
A novel contribution is supporting **supervisor self-awareness**. Supervision is interpretive; feedback is shaped by the supervisor's experience and biases. By pre-conditioning the AI with contextual details, the system analyzes the supervisor’s full body of written feedback. This reflective layer evaluates consistency, identifies over/underemphasized domains, and highlights possible blind spots (e.g., praising scaffolding but neglecting pacing). These patterns are essential for leadership development.

Implications:
Integrating all three phases creates a comprehensive, scalable, and deeply informative AI-first supervision workflow. This shifts the faculty role away from repetitive documentation toward higher-level analysis and coaching, addressing limited supervisory bandwidth. Residents receive clearer, more consistent, evidence-rich feedback. The work offers a replicable AI-first model for integrating generative AI into teacher supervision, transforming qualitative data into actionable insights, reducing faculty overload, and improving feedback reliability. It aligns with INTED’s thematic areas, including Technology-Enhanced Learning, Teacher Training, and Emerging Applications of AI.

Keywords:

Teacher residency, AI workflow, supervision, observation analysis, reflective practice.