# Is ChatGPT a Good Teacher Coach? Measuring Zero-Shot Performance For Scoring and Providing Actionable Insights on Classroom Instruction

**Authors:** Rose E. Wang, Dorottya Demszky  
**arXiv:** [2306.03090](https://arxiv.org/abs/2306.03090)  
**Date:** 2023-06-05  
**Summary page:** https://www.summarizepaper.com/en/paper/2306.03090v1/

## Key points

- Scoring transcript segments based on classroom observation instruments
- Identifying highlights and missed opportunities for good instructional strategies
- Offering actionable suggestions for eliciting more student reasoning
- ChatGPT is able to generate relevant feedback for improving instruction, but lacks novelty and insight
- Collaboration between researchers and instructors is essential to expand generative AI coaching beyond mathematics into other subjects like History or English
- Integrating automated feedback into human coaching practices is viewed as a complement rather than a replacement due to current technological limitations and the importance of human interaction
- Challenges exist in generating insightful feedback, but future research can focus on addressing these obstacles to improve the capacity of generative AI in coaching teachers effectively

## Summary

In this study, we explore the potential of generative AI to serve as an automated teacher coach by proposing three specific tasks for providing feedback to teachers: scoring transcript segments based on classroom observation instruments, identifying highlights and missed opportunities for good instructional strategies, and offering actionable suggestions for eliciting more student reasoning. Our results show that ChatGPT is able to generate relevant feedback for improving instruction; however, it often lacks novelty and insight. This highlights the challenges of producing insightful and truthful feedback for teachers using generative AI. To expand beyond mathematics and into other subjects like History or English, collaboration between researchers and instructors is essential in identifying relevant dimensions of effective teaching. Integrating automated feedback into human coaching practices is seen as a complement rather than a replacement for expert coaching due to current technological limitations and the importance of human interaction in coaching. Overall, this study represents a step towards leveraging generative AI to provide scalable automated feedback on instructional practice. While there are challenges to overcome in generating insightful feedback, future research can focus on addressing these obstacles to improve the capacity of generative AI in coaching teachers effectively.