Elamur Research⌕
☰
home / papers / 2026 / Instructional Agents: Reducing Teaching Faculty Workload through Multi-Agent Instructional Design

Instructional Agents: Reducing Teaching Faculty Workload through Multi-Agent Instructional Design

Paper · 2026 · Huaiyuan Yao, Wanpeng Xu, Justin Turnau et al.

Summary

Building course materials such as syllabi, slides, lecture notes and tests takes faculty a great deal of time, and existing AI tools usually handle only one piece of this work. The authors present Instructional Agents, an LLM-based multi-agent system that produces a full course package: learning objectives, syllabi, LaTeX slides, slide scripts and assessments. Each agent takes on an academic role: a Program Chair, a Course Coordinator, Teaching Faculty, a Teaching Assistant and an Instructional Designer, joined by a simulated Test Student. They work through the Analyze, Design and Develop stages of the ADDIE instructional design model. Four operating modes set how much humans intervene: Autonomous, Catalog-Guided, Feedback-Guided and Full Co-Pilot. The system was tested on five university courses, with both human and automated reviewers judging quality and the editing effort faculty would still need. Fully autonomous runs are the cheapest and fastest. Human involvement, especially in Co-Pilot mode, gives the best materials. Ablations show that splitting the work across specialized roles clearly beats a single model. The authors argue this matters most for institutions that lack dedicated instructional design staff.

Key points

  • Ablation scores (average quality):
    • Single GPT-4o-mini agent: 2.33.
    • Full system in autonomous mode: 3.01.
    • Full Co-Pilot mode: 3.74.
    • Removing Teaching Faculty, Teaching Assistant or Instructional Designer lowered the average to 2.75, 2.91 and 2.80 respectively.
  • Effect of individual roles:
    • Without the Instructional Designer, learning objectives and syllabi suffered most.
    • Without Teaching Faculty, syllabi and slides suffered most.
  • Generation success rates:
    • gpt-4o: 98.88%.
    • gpt-4o-mini: 98.36%.
    • o1-preview: 97.78%.
    • Nearly all failures were slide LaTeX that pdflatex could not compile.
  • Scope: Only the first three ADDIE phases are covered. Implement and Evaluate are left out so that humans review materials before students see them.
  • Limitations:
    • Visual and interactive content is weak.
    • Feedback is applied by regenerating material rather than through targeted edits.
    • Bias was checked only in an auxiliary CEAT analysis.
    • Code is released via the project website.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Arizona State UniversityTempe · United States50 works · 2,032 citations

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksInformation Systems · 1,437 works

Cite

Cite as Huaiyuan Yao, Wanpeng Xu, Justin Turnau, Nadia Kellam and Hua Wei. 2026. Instructional Agents: Reducing Teaching Faculty Workload through Multi-Agent Instructional Design. Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.eacl-long.191

@inproceedings{yao2026instructional,
  title = {{Instructional Agents: Reducing Teaching Faculty Workload through Multi-Agent Instructional Design}},
  author = {Huaiyuan Yao, Wanpeng Xu, Justin Turnau, Nadia Kellam and Hua Wei},
  booktitle = {Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.eacl-long.191},
  url = {https://doi.org/10.18653/v1/2026.eacl-long.191}
}

References

37 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Maria Ijaz Baig and Elaheh Yadegaridehkordi. 2024. Chatgpt in the higher education: A systematic literature review and research challenges. International Journal of Educational Research, 127:102411. find
  2. Sue Bennett, Shirley Agostinho, and Lori Lockyer. 2017. The process of designing for learning: Understanding university teachers’ design work. Educational Technology Research and Development, 65:125–145. find
  3. John Biggs. 1996. Enhancing teaching through constructive alignment. Higher education, 32(3):347– 364. find
  4. John Biggs, Catherine Tang, and Gregor Kennedy. 2022. Teaching for quality learning at university 5e. McGraw-hill education (UK). find
  5. Robert Maribe Branch and İlhan Varank. 2009. Instructional design: The ADDIE approach, volume 722. Springer. find
  6. Sara E Brownell and Kimberly D Tanner. 2012. Barriers to faculty pedagogical change: Lack of training, time, incentives, and. . . tensions with professional identity? CBE—Life Sciences Education, 11(4):339– 346. find
  7. Zhendong Chu, Shen Wang, Jian Xie, Tinghui Zhu, Yibo Yan, Jinheng Ye, Aoxiao Zhong, Xuming Hu, Jing Liang, Philip S Yu, et al. 2025. Llm agents for education: Advances and applications. arXiv preprint arXiv:2503.11733. find
  8. Robert O Davis and Yong Jik Lee. 2023. Prompt: Chatgpt, create my course, please! Education Sciences, 14(1):24. find
  9. Haoxiang Fan, Guanzheng Chen, Xingbo Wang, and Zhenhui Peng. 2024. Lessonplanner: Assisting novice teachers to prepare pedagogy-driven lesson plans with large language models. In Proceedings of the 37th Annual ACM Symposium on User Interface Software and Technology, pages 1–20. find
  10. Milton Friedman. 1937. The use of ranks to avoid the assumption of normality implicit in the analysis of variance. Journal of the American Statistical Association, 32(200):675–701. find
  11. Robert M Gagne, Walter W Wager, Katharine C Golas, John M Keller, and James D Russell. 2005. Principles of instructional design. find
  12. Mihajla Gavin and Susan McGrath-Champ. 2024. Teacher workload and the organisation of work: a research agenda for a post-pandemic future. Labour and Industry, 34(1):88–99. find
  13. Zhanxin Hao, Fei Qin, Jianxiao Jiang, Jie Cao, Jifan Yu, Zhiyuan Liu, and Yu Zhang. 2025. Ai as learning partners: Students’ interactions and perceptions in a simulated classroom with multiple llm-powered agents. In Proceedings of the 19th International Conference of the Learning Sciences-ICLS 2025,… find
  14. Yihan Hong, Huaiyuan Yao, Bolin Shen, Wanpeng Xu, Hua Wei, and Yushun Dong. 2026. Rulers: Locked rubrics and evidence-anchored scoring for robust llm evaluation. arXiv preprint arXiv:2601.08654. — [arXiv:2601.08654](https://arxiv.org/abs/2601.08654) find
  15. Bihao Hu, Jiayi Zhu, Yiying Pei, and Xiaoqing Gu. 2025. Exploring the potential of llm to enhance teaching plans through teaching simulation. npj Science of Learning, 10(1):7. find
  16. Enkelejda Kasneci, Kathrin Seßler, Stefan Küchemann, Maria Bannert, Daryna Dementieva, Frank Fischer, Urs Gasser, Georg Groh, Stephan Günnemann, Eyke Hüllermeier, et al. 2023. Chatgpt for good? on opportunities and challenges of large language models for education. Learning and individual difference…
  17. Pingchuan Ma, Tsun-Hsuan Wang, Minghao Guo, Zhiqing Sun, Joshua B Tenenbaum, Daniela Rus, Chuang Gan, and Wojciech Matusik. 2024. Llm and simulation as bilevel optimizers: A new paradigm to advance physical scientific discovery. arXiv preprint arXiv:2405.09783. find
  18. Eileen G Merritt. 2016. Time for teacher learning, planning critical for school reform. Phi delta kappan, 98(4):31–36. find
  19. OpenAI. 2024a. Gpt-4o model card. https://platform.openai.com/docs/models/gpt-4o. ↗
  20. OpenAI. 2024b. Introducing gpt-4o-mini. https://platform.openai.com/docs/models/ gpt-4o-mini. Accessed: 2025-08-01. ↗
  21. OpenAI. 2024c. Introducing o1-preview. https://platform.openai.com/docs/models/ o1-preview. Accessed: 2025-08-01. ↗
  22. Joon Sung Park, Joseph O’Brien, Carrie Jun Cai, Meredith Ringel Morris, Percy Liang, and Michael S Bernstein. 2023. Generative agents: Interactive simulacra of human behavior. In Proceedings of the 36th annual acm symposium on user interface software and technology, pages 1–22. find
  23. Jingyang Peng, Wenyuan Shen, Jiarui Rao, and Jionghao Lin. 2025. Automated bias assessment in ai-generated educational content using ceat framework. Preprint, arXiv:2505.12718. — [arXiv:2505.12718](https://arxiv.org/abs/2505.12718) find
  24. Quality Matters. 2023. Higher Education Rubric, Seventh Edition. MarylandOnline, Inc. Available at https://www.qualitymatters.org/sites/default/files/PDFs/ StandardsfromtheQMHigherEducationRubric. pdf. ↗
  25. Sumedh Rasal and EJ Hauer. 2024. Navigating complexity: Orchestrated problem solving with multi-agent llms. arXiv preprint arXiv:2402.16713. — [arXiv:2402.16713](https://arxiv.org/abs/2402.16713) find
  26. Sam Toorchi Roodsari and Shahram Azizi Ghanbari. 2024. Instructional design and ai in learning environments: Developing competency-validated adaptive feedback for higher education. In London International Conference On Education. find
  27. Lina Zhao, Jiaxing Bai, Zihao Bian, Qingyue Chen, Yafang Li, Guangbo Li, Min He, Huaiyuan Yao, and Zongjiu Zhang. 2025. Autonomous multi-modal llm agents for treatment planning in focused ultrasound ablation surgery. Preprint, arXiv:2505.21418. — [arXiv:2505.21418](https://arxiv.org/abs/2505.21418) find
  28. Samuel Schmidgall, Yusheng Su, Ze Wang, Ximeng Sun, Jialian Wu, Xiaodong Yu, Jiang Liu, Zicheng Liu, and Emad Barsoum. 2025. Agent laboratory: Using llm agents as research assistants. https://arxiv.org/abs/2501.04227. ↗
  29. Shen Wang, Tianlong Xu, Hang Li, Chaoli Zhang, Joleen Liang, Jiliang Tang, Philip S. Yu, and Qingsong Wen. 2024. Large language models for education: A survey and outlook. ArXiv, abs/2403.18105. find
  30. Xiaoyan Wang, Yelin Su, Stephen Cheung, Eva Wong, and Theresa Kwong. 2013. An exploration of biggs’ constructive alignment in course design and its impact on students’ learning approaches. Assessment & Evaluation in Higher Education, 38(4):477–491. find
  31. Yaoxiang Wang, Zhiyong Wu, Junfeng Yao, and Jinsong Su. 2025. Tdag: A multi-agent framework based on dynamic task decomposition and agent generation. Neural Networks, page 107200. find
  32. Ke Yang, Yao Liu, Sapana Chaudhary, Rasool Fakoor, Pratik Chaudhari, George Karypis, and Huzefa Rangwala. 2024. Agentoccam: A simple yet strong baseline for llm-based web agents. arXiv preprint arXiv:2410.13825. — [arXiv:2410.13825](https://arxiv.org/abs/2410.13825) find
  33. Huaiyuan Yao, Longchao Da, Vishnu Nandam, Justin Turnau, Zhiwei Liu, Linsey Pang, and Hua Wei. 2025a. Comal: Collaborative multi-agent large language models for mixed-autonomy traffic. In Proceedings of the 2025 SIAM International Conference on Data Mining (SDM), pages 409–418. SIAM. find
  34. Huaiyuan Yao, Pengfei Li, Bu Jin, Yupeng Zheng, An Liu, Lisen Mu, Qing Su, Qian Zhang, Yilun Chen, and Peng Li. 2025b. Lilodriver: A lifelong learning framework for closed-loop motion planning in long-tail autonomous driving scenarios. Preprint, arXiv:2505.17209. — [arXiv:2505.17209](https://arxiv.org/abs/2505.17209) find
  35. Olaf Zawacki-Richter, John YH Bai, Kyungmee Lee, Patricia J Slagter van Tryon, and Paul Prinsloo. 2024. New advances in artificial intelligence applications in higher education? International Journal of Educational Technology in Higher Education, 21(1):32. find
  36. Xiaoming Zhai. 2023. Chatgpt for next generation science learning. XRDS: Crossroads, The ACM Magazine for Students, 29(3):42–46. find
  37. Zheyuan Zhang, Daniel Zhang-Li, Jifan Yu, Linlu Gong, Jinchang Zhou, Zhanxin Hao, Jianxiao Jiang, Jie Cao, Huiqin Liu, Zhiyuan Liu, et al. 2024. Simulating classroom education with llm-empowered agents. arXiv preprint arXiv:2406.19226.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/yao-2026-instructional-agents-reducing-teaching-faculty/