AI-driven and Closed-loop Human Tutor Training and Real-life Assessment

There exists a critical disconnect between human tutor professional development and actual student learning. In this talk, I will present a novel, AI-driven system designed to "close the loop" between human tutor training, real-world pedagogical actions, and student learning outcomes.

I will discuss a "learning by doing" approach where tutors practice specific competencies via AI-enhanced scenarios. Moving beyond simple training metrics, I will demonstrate how we utilize Large Language Models (Gemini-2.5-pro) to analyze transcripts of authentic, real-life tutoring sessions. We will explore the data science techniques used to validate this pipeline—specifically, how we measure the transfer of skills from tutor training to practice and how we establish the internal validity linking those specific "tutor moves" to demonstrable gains in student learning. This session will offer a roadmap for using AI to rigorously validate the impact of workforce training on the ultimate goal: student success.