Journal of Environmental Impact Assessment (환경영향평가)

Korean Society of Environmental Impact Assessment (한국환경영향평가학회)
권호 표지
DOI QR코드

DOI QR Code

Pilot Application and Expansion Direction of Generative AI for the Review of Environmental Impact Assessment Reports

환경영향평가 검토지원을 위한 생성형 AI의 시범적용과 확대 방향

  • Received : 2024.09.27
  • Accepted : 2024.10.10
  • Published : 2024.10.31
Download PDF
⟨ Previous Next ⟩

Abstract

The review of environmental impact assessment reports is a crucial process to ensure the expertise and objectivity of the content and procedures of Environmental Impact Assessment (EIA). Since the amendment of the Environmental Impact Assessment Act in 1997, a single national research institute has independently and exclusively supported this role based on legal grounds. In 2021, in an effort to secure a more diverse range of review opinions, there was a movement towards decentralization and diversification of review institutions. However, concerns have been raised about the limitations in the independence and consistency of these review opinions. Meanwhile, with the introduction of OpenAI's ChatGPT in November 2022, the use of generative artificial intelligence (AI) has increased, becoming common in daily life, work, and research. This has led to discussions about the potential use of generative AI to leverage existing EIA review capabilities and ensure a balanced range of opinions from the recently decentralized and diversified institutions. This study aims to propose a pilot implementation and future methodological and institutional expansion of generative AI to support the review of environmental impact assessment reports. To achieve this, a generative AI system was initially developed and tested to support the review of content related to Health Impact Assessment within the EIA framework. The practical applications of this pilot system were documented. Additionally, by specifying the procedures, content, and stakeholders involved in the EIA review process, this paper suggests considerations for its expansion from both methodological and institutional perspectives.

환경영향평가서 검토는 환경영향평가 내용 및 절차의 전문성, 객관성 등을 확보하기 위한 중요한 과정으로, 1997년 환경영향평가법 개정 이후 최근까지 법적 근거에 따라 단일의 국책연구기관이 독립적이며 독점적으로 그 역할을 지원해 왔다. 이와 관련하여 2021년에는 다양한 검토의견 확보 차원에서 검토기관의 분산화 및 다원화가 진행되었으나, 검토의견의 독립성 및 일관성에 한계가 상존한다는 의견이 제기되었다. 다른 한편으로 2022년 11월에 OpenAI의 ChatGPT가 소개되면서 생성형 AI의 활용이 증가해 왔으며 최근에는 일상, 업무, 그리고 연구에까지 보편화되고 있다. 이에 그간의 환경영향평가서 검토 역량 활용과 최근 분산/다원화된 기관들의 검토의견 균형성 확보 목적으로 생성형 AI의 활용 가능성이 언급되고 있다. 본 논문에서는 환경영향평가서 검토지원을 위한 생성형 AI의 시범적인 구축과 향후 방법론적 제도적 확대 방향을 제안하고자 하였다. 이를 위하여 우선 환경영향평가 제도 내에서 운영되는 건강영향평가 관련 내용의 검토지원을 위한 생성형 AI 시스템을 시범적으로 구축하여 활용 사항을 정리하였다. 이와 더불어 환경영향평가서 검토의 절차, 내용, 이해관계자 등을 구체화하여 방법론적 측면과 제도적 측면에서 도입 시 고려사항을 검토하여 제안하였다.

Keywords

Acknowledgement

본 논문은 환경부의 재원으로 한국환경산업기술원의 환경보건디지털조사 기반구축 개발사업(과제번호: 2021-KE001620)의 지원을 받아 2024년 한국환경연구원에서 수행한 「환경유해인자 노출에 의한 공간의 환경보건상태 평가 기술개발(2024-005(R))」 및 한국환경연구원 연구개발적립금 사업인 「초거대 언어모델을 활용한 지능형 환경지식 검색시스템 구축(RR2024-16)」 사업의 연구결과 일부를 바탕으로 작성되었습니다.

References

  1. Al Ghadban Y, Lu HY, Adavi U, Sharma A, Gara S, Das N, Kumar B, John R, Devarsetty P, Hirst JE. 2024, Transforming Healthcare Education: Harnessing Large Language Models for Frontline Health Worker Capacity Building using Retrieval-Augmented Generation. In: NeurIPS'23 Workshop on Generative AI for Education (GAIED); 2023 Dec 15; New Orleans (USA); Ernest N. Morial Convention Center. Available from: https://gaied.org/neurips2023/files/30/30_paper.pdf 
  2. Chiang W-L, Zheng L, Sheng Y, Angelopoulos AN, Li T, Li D, Zhang H, Zhu B, Jordan M, Gonzalez JE, Stoica I. 2024. Chatbot Arena: An Open Platform for Evaluating LLMs by Human Preference. arXiv e-prints. [accessed 2024 Aug 20]:[29 p.]. Available from: https://doi.org/10.48550/arXiv.2403.04132 
  3. Douze M, Guzhva A, Deng C, Johnson J, Szilvasy G, Mazare P-E, Lomeli M, Hosseini L, Jegou H. 2024. The Faiss library. arXiv e-prints, [accessed 2024 Jun 25]:[22 p.]. Available from: https://doi.org/10.48550/arXiv.2401.08281 
  4. Edge D, Trinh H, Cheng N, Bradley J, Chao A, Mody A, Truitt S, Larson J. 2024. From Local to Global: A Graph RAG Approach to Query-Focused Summarization. arXiv e-prints. [accessed 2024 Jun 18]:[15 p.]. Available from: https://arxiv.org/abs/2404.16130 
  5. Gao Y, Xiong Y, Gao X, Jia K, Pan J, Bi Y, Dai Y, Sun J, Wang M, Wang H. 2024. Retrieval-Augmented Generation for Large Language Models: A Survey. arXiv e-prints, [accessed 2024 Jul 19]:[21 p.]. Available from: https://doi.org/10.48550/arXiv.2312.10997 
  6. Heimerl F, Gleicher M. 2018. Interactive Analysis of Word Vector Embeddings. Compute. Graph. Forum, 37(3): 253-265. Available from: https://doi.org/10.1111/cgf.13417 
  7. Hugging Face. 2024. Hugging Face. [accessed 2024 Aug 10]. Available from: https://huggingface.co/
  8. Jablonka, KM, Schwaller P, Ortega-Guerrero A, Smit B. 2024. Leveraging large language models for predictive chemistry. Nat. Mach. Intell. 6(2):161-169. Available from: https://doi.org/10.1038/s42256-023-00788-1 
  9. Jeon, J. and Lee, S. 2023. Large language models in education: A focus on the complementary relationship between human teachers and ChatGPT. Educ. Inf. Technol. 28(12):15873-15892. Available from: https://doi.org/10.1007/s10639-023-11834-1 
  10. LangChain. 2024. RecursiveCharacterTextSplitter [Internet]; [accessed 2024 Jul 29]. Available from: https://python.langchain.com/v0.2/api_reference/text_splitters/character/langchain_text_splitters.character.RecursiveCharacterTextSplitter.html 
  11. LangChain Hub. 2024. rlm/rag-prompt [Internet]; [accessed 2024 Jul 29]. Available from: https://smith.langchain.com/hub/rlm/rag-prompt 
  12. Liu Y, He H, Han T, Zhang X, Liu M, Tian J, Zhang Y, Wang J, Gao X, Zhong T, Pan Y, Xu S, Wu Z, Liu Z, Zhang X, Zhang S, Hu X, Zhang T, Qiang N, Liu T, Ge B. 2024. Understanding LLMs: A comprehensive overview from training to inference. arXiv e-prints, [accessed 2024 Aug 20]:[30 p.]. Available from: https://doi.org/10.48550/arXiv.2401.02038 
  13. Madani A, Krause B, Greene E, Subramanian S, Mohr B, Holton J, Olmos J, Xiong C, Sun Z, Socher R, Fraser J, and Naik N. 2023. Large language models generate functional protein sequences across diverse families. Nat. Biotechnol. 41(8):1099-1106. Available from: https://doi.org/10.1038/s41587-022-01618-2 
  14. Ministry of Environment. 2020. A study on the improvement of health impact assessment for consumers (II).
  15. Ministry of Environment. 2023. Manual for Additional Assessment of Health Impact Items.
  16. Ministry of Environment. 2024a. Compilation of Regulations Related to Environmental Impact Assessment.
  17. Ministry of Environment. 2024b. Guidelines for the Preparation of Environmental Impact Statements.
  18. National Law Information Center. 2024a. Environmental Impact Assessment Act. Act No. 20334, Partially Amended on February 20, 2024. [accessed 2024 Aug 13]. Available from: https://law.go.kr/%EB%B2%95%EB%A0%B9/%ED%99%98%EA%B2%BD%EC%98%81%ED%96%A5%ED%8F%89%EA%B0%80%EB%B2%95
  19. National Law Information Center. 2024b. Enforcement Decree of the Environmental Impact Assessment Act. Presidential Decree No. 34656, Amended by Other Laws, Article 23, Paragraph 2. [accessed 2024 Aug 13]. Available from: https://law.go.kr/LSW//lsInfoP.do?lsiSeq=263619&ancYd=20240702&ancNo=34656&efYd=20240710&nwJoYnInfo=Y&efGubun=Y&chrClsCd=010202&ancYnChk=0#0000
  20. National Law Information Center. 2024c. Environmental Health Act. Act No. 19668, Partially Amended on August 16, 2023, Article 13. [accessed 2024 Aug 13]. Available from: https://www.law.go.kr/lsSc.do?section=&menuId=1&subMenuId=15&tabMenuId=81&eventGubun=060101&query=%ED%99%98%EA%B2%BD%EB%B3%B4%EA%B1%B4%EB%B2%95#undefined
  21. National Law Information Center. 2024d. Regulations on the Processing of Consultation Tasks for Environmental Impact Statements, etc. Partially Amended on June 19, 2024, Article 13. [accessed 2024 Aug 13]. Available from: https://law.go.kr/LSW/admRulLsInfoP.do?admRulId=25874&efYd=0
  22. National Law Information Center. 2024e. Regulations on the preparation of environmental impact statements, Partially Amended on April 13, 2023, Article 13. [accessed 2024 Aug 13]. Available from: https://law.go.kr/LSW/admRulLsInfoP.do?admRulId=26073&efYd=0
  23. OpenAI. 2024a. New embedding models and API updates; [accessed 2024 Jul 29]. Available from: https://openai.com/index/new-embedding-models-and-api-updates/
  24. OpenAI. 2024b. Tokenizer; [accessed 2024 Jul 29]. Available from:https://platform.openai.com/tokenizer 
  25. Patil R, Gudivada V. 2024. A review of current trends, techniques, and challenges in large language models (LLMs). Appl. Sci. 14(5): 2074. Available from: https://doi.org/10.3390/app14052074 
  26. Ro TH. 2021. Diversification of Environmental Impact Assessment Review Agencies: Challenges and Prospects. Korea Environment Institute Focus. 9(7):1-16, Available from: https://www.kei.re.kr/board.es?mid=a10102020000&bid=0028&act=view&list_no=57621 [Korean Literature]  10102020000&bid=0028&act=view&list_no=57621
  27. Sallam, M. 2023. ChatGPT Utility in Healthcare Education, Research, and Practice: Systematic Review on the Promising Perspectives and Valid Concerns. Healthcare, 11(6): 887. Available from: https://doi.org/10.3390/healthcare11060887 
  28. Setty S, Thakkar H, Lee A, Chung E, Vidra N. 2024. Improving Retrieval for RAG based Question Answering Models on Financial Documents. arXiv e-prints. [accessed 2024 Aug 10]:[14 p.]. Available from: https://arxiv.org/abs/2404.07221 
  29. Snowflake Inc. 2024, Streamlit [software]. Ver. 1.37.0. Python Package Index. [accessed 2024 Jul 29]. Available from: https://pypi.org/project/streamlit/1.37.0/#files 
  30. Tian S, Jin Q, Yeganova L, Lai P.-T., Zhu Q, Chen X, Yang Y, Chen Q, Kim W, Comeau DC, Islamaj R, Kapoor A, Gao X, Lu Z. 2024. Opportunities and challenges for ChatGPT and large language models in biomedicine and health. Briefings in Bioinformatics, 25(1):bbad493. Available from: https://doi.org/10.1093/bib/bbad493 
  31. Vaswani A, Shazeer N, Parmar N, Uszkoreit J, Jones L, Gomez AN, Kaiser L, Polosukhin I. 2017. Attention is all you need. In Proceedings of the 31st International Conference on Neural Information Processing Systems (NIPS'17). Curran Associates Inc., Red Hook, NY, USA, 6000-6010. 
  32. Zhang T, Patil S, Jain N, Shen S, Zaharia M, Stoica I, Gonzalez J. 2024, RAFT: Adapting Language Model to Domain Specific RAG, arXiv e-prints. [accessed 2024 Aug 20]:[12 p.]. Available from: https://doi.org/10.48550/arXiv.2403.10131