AI Feedback for LEED Narratives: Exploring LLM Potential

11 pages•Published: July 23, 2025

Zhenlin Yang, Laura M. Cruz Castro and Gabriel Castelblanco

Abstract

Effective feedback is crucial in sustainability design education, where students must meet complex criteria such as Leadership in Energy and Environmental Design (LEED) standards. This study explores the potential for an automated feedback system to be powered by Large Language Models (LLMs), designed to deliver real-time, detailed, and rubric-aligned feedback for sustainability-focused projects. Through quantitative and qualitative analysis, the tool’s performance was compared to teaching assistant (TA) feedback, focusing on feedback length, alignment with LEED rubrics, clarity, and timeliness. The results show that Llama 8B 3.1 has significantly longer and more structured feedback, offering immediate and iterative responses that facilitated systematic refinement of student work. In contrast, TA feedback, while shorter and often delayed, provided nuanced, context-sensitive insights tailored to individual submissions. The findings demonstrate the effectiveness in delivering standardized and scalable feedback by Llama 8B 3.1, while also highlighting the value of TAs' personalized guidance. This study advances the field of automated feedback systems and underscores the potential of integrating AI-powered tools with human expertise to enhance sustainability design education.

Keyphrases: feedback, leed, llama

In: Wesley Collins, Anthony J. Perrenoud and John Posillico (editors). Proceedings of Associated Schools of Construction 61st Annual International Conference, vol 6, pages 321-331.

Links:	https://easychair.org/publications/paper/Kb1w
	https://doi.org/10.29007/mgfs

BibTeX entry

@inproceedings{ASC2025:AI_Feedback_LEED_Narratives,
  author    = {Zhenlin Yang and Laura M. Cruz Castro and Gabriel Castelblanco},
  title     = {AI Feedback for LEED Narratives: Exploring LLM Potential},
  booktitle = {Proceedings of Associated Schools of Construction 61st Annual International Conference},
  editor    = {Wesley Collins and Anthony J. Perrenoud and John Posillico},
  series    = {EPiC Series in Built Environment},
  volume    = {6},
  publisher = {EasyChair},
  bibsource = {EasyChair, https://easychair.org},
  issn      = {2632-881X},
  url       = {/publications/paper/Kb1w},
  doi       = {10.29007/mgfs},
  pages     = {321-331},
  year      = {2025}}

Download PDF Open PDF in browser