HUMAN-AI CO-CREATION SYSTEM FOR KNOWLEDGE WORK BASED ON MULTI-AGENT APPROACH
Keywords:
Human–AI collaboration, Generative AI, Large Language Model, Intelligent systemAbstract
The task of writing a scientific research proposal is complex and highly structured, yet traditional writing methods are typically inefficient. To address this challenge, this study introduces an intelligent writing system leveraging a large language model (LLM). The core contribution is a modular proposal drafting framework that automatically generates multi-chapter application content based on user-specified disciplines and topic questions. The system first creates outlines and content overviews for each chapter through intent recognition, then composes detailed chapter content guided by these outlines. Finally, the system consolidates these components into a complete proposal draft. This modular architecture not only ensures logical consistency across the document but also empowers users to independently refine and optimize individual chapters. To evaluate the system's efficacy, we invited multiple researchers for assessment. Benchmarking against a single-agent LLM demonstrates that our multi-agent system produces proposals with superior content coverage, logical coherence, and user satisfaction, while significantly improving writing efficiency. The proposed modular prompt framework exhibits broader applicability and can be readily extended to other funding application contexts, offering a novel technological approach for advancing intelligent research support systems.References
[1] Lindgreen A, Di Benedetto CA, Verdich C, et al. How to write really good research funding applications. Industrial Marketing Management, 2019, 77: 232-239.
[2] Ren R, Ma J, Zheng Z. Large language model for interpreting research policy using adaptive two-stage retrieval augmented fine-tuning method. Expert Systems with Applications, 2025, 278: 127330.
[3] Locke LF, Spirduso WW, Silverman SJ. Proposals that work: A guide for planning dissertations and grant proposals. Sage Publications, 2013.
[4] Herbert DL, Barnett AG, Clarke P, et al. On the time spent preparing grant proposals: an observational study of Australian researchers. BMJ Open, 2013, 3(5): e002800.
[5] Russo F. Automated content writing tools and the question of objectivity. Digital Society, 2023, 2(3): 50.
[6] Wu J, Yang S, Zhan R, et al. A survey on LLM-generated text detection: Necessity, methods, and future directions. Computational Linguistics, 2025: 1-66.
[7] Ren R, Ma J, Luo J. Large language model for patent concept generation. Advanced Engineering Informatics, 2025, 65: 103301.
[8] Wang Y, Guo Q, Yao W, et al. Autosurvey: Large language models can automatically write surveys. Advances in Neural Information Processing Systems, 2024, 37: 115119-115145.
[9] Kallet RH. How to write the methods section of a research paper. Respiratory Care, 2004, 49(10): 1229-1232.
[10] Sharma R, Gulati S, Kaur A, et al. Research discovery and visualization using ResearchRabbit: A use case of AI in libraries. COLLNET Journal of Scientometrics and Information Management, 2022, 16(2): 215-237.
[11] Zhou YC, Zheng Z, Lin JR, et al. Integrating NLP and context-free grammar for complex rule interpretation towards automated compliance checking. Computers in Industry, 2022, 142: 103746.
[12] Pal S, Bhattacharya M, Islam MA, et al. AI-enabled ChatGPT or LLM: A new algorithm is required for plagiarism-free scientific writing. International Journal of Surgery, 2024, 110(2): 1329-1330.
[13] Mohammad S, Dorr B, Egan M, et al. Using citations to generate surveys of scientific paradigms. Proceedings of Human Language Technologies: The 2009 Annual Conference of the North American Chapter of the Association for Computational Linguistics, 2009, June: 584-592.
[14] Jha R, Finegan-Dollak C, King B, et al. Content models for survey generation: A factoid-based evaluation. Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), 2015, July: 441-450.
[15] Sun X, Zhuge H. Automatic generation of survey paper based on template tree. 2019 15th International Conference on Semantics, Knowledge and Grids (SKG), 2019, September: 89-96.
[16] Liu S, Cao J, Yang R, Wen Z. Generating a structured summary of numerous academic papers: Dataset and method. International Joint Conferences on Artificial Intelligence, 2022.
[17] Zhu K, Feng X, Feng X, et al. Hierarchical catalogue generation for literature review: A benchmark. Findings of the Association for Computational Linguistics: EMNLP 2023, 2023, December: 6790-6804.
[18] Kaneko M, Mita M, Kiyono S, et al. Encoder-decoder models can benefit from pre-trained masked language models in grammatical error correction. arXiv preprint, arXiv:2005.00987, 2020.
[19] Omelianchuk K, Atrasevych V, Chernodub A, et al. GECToR–Grammatical error correction: Tag, not rewrite. Proceedings of the Fifteenth Workshop on Innovative Use of NLP for Building Educational Applications, 2020, July: 163-170.
[20] Mitchell P, Riedlinger M, Goldenfein J, et al. Research GenAI: Situating generative AI in the scholarly economy. AoIR Selected Papers of Internet Research, 2024.
[21] Kinnunen T, Leisma H, Machunik M, et al. SWAN-scientific writing AssistaNt: A tool for helping scholars to write reader-friendly manuscripts. Proceedings of the Demonstrations at the 13th Conference of the European Chapter of the Association for Computational Linguistics, 2012, April: 20-24.
[22] Adhi P. Exploring the use of ChatGPT as a supporting tool in writing research proposals: EFL students' perspectives. Doctoral dissertation, UIN Sunan Gunung Djati Bandung, 2024.
[23] Bai X, Stede M. A survey of current machine learning approaches to student free-text evaluation for intelligent tutoring. International Journal of Artificial Intelligence in Education, 2023, 33(4): 992-1030.
[24] Brown T, Mann B, Ryder N, et al. Language models are few-shot learners. Advances in Neural Information Processing Systems, 2020, 33: 1877-1901.
[25] Wang Q, Huang L, Jiang Z, et al. PaperRobot: Incremental draft generation of scientific ideas. arXiv preprint, arXiv:1905.07870, 2019.
[26] Taylor R, Kardas M, Cucurull G, et al. Galactica: A large language model for science. arXiv preprint, arXiv:2211.09085, 2022.
[27] Huang J, Tan M. The role of ChatGPT in scientific communication: Writing better scientific review articles. American Journal of Cancer Research, 2023, 13(4): 1148.
[28] Seckel E, Stephens BY, Rodriguez F. Ten simple rules to leverage large language models for getting grants. PLOS Computational Biology, 2024, 20(3): e1011863.
[29] Wang Y, Guo Q, Yao W, et al. Autosurvey: Large language models can automatically write surveys. Advances in Neural Information Processing Systems, 2024, 37: 115119-115145.
[30] Du Z, Hashimoto K. AcademiCraft: Transforming writing assistance for English for academic purposes with multi-agent system innovations. Information, 2025, 16(4).
[31] Daudaravicius V. Automated evaluation of scientific writing: AESW shared task proposal