Development of Generative AI Output Control Algorithm for Disaster and Safety Management Legislation Utilization

doi:10.15683/kosdi.2025.12.31.1027

All Issue

2025 Vol.21, Issue 4 Preview Page Next Page

Research Article

Development of Generative AI Output Control Algorithm for Disaster and Safety Management Legislation Utilization 재난 안전 관리 법령 활용을 위한 생성형 AI 출력 제어 알고리즘 개발: 배종옥¹ㆍ임준묵²*
Jong-ok Bae¹, Joon-mook Lim²*; ¹Graduate Student, Department of Creative Convergence Engineering, Hanbat National University, Daejeon, Republic of Korea
²Professor, Department of Creative Convergence Engineering, Hanbat National University, Daejeon, Republic of Korea

31 December 2025. pp. 1027-1044

PDF

Abstract

Purpose: This study addresses the limitation of existing generative AI systems that rely on single-output responses and cannot fully support the disaster management cycle of prevention, preparedness, response, and recovery. To overcome this limitation, a multi-channel generative AI output control algorithm that reflects all four stages simultaneously for a single query is proposed. Method: The proposed algorithm is based on a Retrieval-Augmented Generation (RAG) architecture integrated with a database of 5,485 domestic legal provisions in JSON format. Relevant legal articles are retrieved using FAISS-based vector search and ELECTRA embeddings, and four parallel responses are generated using GPT-4. Result: Evaluation using 400 test queries achieved an overall accuracy of 82.8%. Stage-specific F1 scores were 0.780 for prevention, 0.822 for preparedness, 0.825 for response, and 0.879 for recovery, with an average response time of 3.8 seconds. Conclusion: The proposed four-channel output control algorithm improves the reliability and multidimensional decision-making capability of generative AI systems and demonstrates applicability beyond disaster safety management to other domains requiring legally grounded AI responses.

Keywords

Generative Artificial Intelligence

Disaster Management

Multi-Channel Output Control

Legally Grounded AI

RAG Architecture

연구목적: 본 연구는 기존 생성형 인공지능이 단일출력 방식에 의존함으로써 재난관리의 예방·대비·대응·복구 전 주기를 충분히 지원하지 못하는 한계를 해결하고자, 하나의 질의에 대해 네 단계를 동시에 반영하는 다채널 생성형 AI 출력 제어 알고리즘을 제안한다. 연구방법: 제안된 알고리즘은 RAG 아키텍처를 기반으로 5,485개의 국내 법령 JSON 데이터베이스와 연동되며, FAISS 기반 벡터 검색과 ELECTRA 임베딩을 활용하여 관련 법령 조문을 검색하고 GPT-4를 통해 4채널 병렬 응답을 생성한다. 연구결과: 400개의 테스트 질문을 활용한 평가 결과 전체 정확도는 82.8%로 나타났으며, 예방·대비·대응·복구 단계별 F1 점수는 각각 0.780, 0.822, 0.825, 0.879를 기록하였다. 평균 응답 시간은 3.8초였다. 결론: 본 연구의 4채널 출력 제어 알고리즘은 생성형 AI의 출력 신뢰성과 다차원 의사결정 지원 능력을 향상시키며, 재난안전관리뿐 아니라 법적 근거 기반 AI 응답이 요구되는 다양한 분야로의 적용 가능성을 제시한다.

키워드

생성형 인공지능

재난관리

다채널 출력 제어

법적 근거 기반 AI

RAG 아키텍처

References

Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J.D., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., Agarwal, S., Herbert-Voss, A., Krueger, G., Henighan, T., Child, R., Ramesh, A., Ziegler, D.M., Wu, J., Winter, C., Hesse, C., Chen, M., Sigler, E., Litwin, M., Gray, S., Chess, B., Clark, J., Berner, C., McCandlish, S., Radford, A., Sutskever, I., Amodei, D. (2020). “Language models are few-shot learners.” Advances in Neural Information Processing Systems, Vol. 33, pp. 1877-1901.
Choi, H.-C. (2010). “A critical review of the conventional four-phase disaster-management model: Focusing on the response and recovery stages.” Korean Journal of Crisis and Emergency Management, Vol. 6, No. 1, pp. 201-218.
Choi, W. (2020). “Building an intelligent disaster-safety management system based on artificial intelligence.” Journal of the Korean Society of Hazard Mitigation, Vol. 20, No. 1, pp. 127-135. 10.9798/KOSHAM.2020.20.1.127
Chung, K.-S., Jung, W.-S. (2023). “A study on the dataset structure of digital twin for disaster and safety management.” Journal of the Institute of Internet, Broadcasting and Communication, Vol. 23, No. 5, pp. 89-95. 10.7236/JIIBC.2023.23.5.89
Jung, G.S., Jung, W.S. (2022). “Design of social disaster safety platform based on structured/unstructured data.” Journal of the Korea Society of Disaster Information, Vol. 18, No. 4, pp. 715-724.
Kim, J. D. (2010). “An analysis on the order of priority in disaster management policy.” Journal of the Korean Society of Hazard Mitigation, Vol. 10, No. 2, pp. 61-68.
Kim, J., Kim, D.J., Sohn, H.G., Choi, J., Im, J. (2022). “Remote sensing and GIS-based disaster prediction, monitoring, and response (Editorial).” Korean Journal of Remote Sensing, Vol. 38, No. 5-2, pp. 661-667.
Kim, M., Kim, G.Y., Jang, D., Yang, M.S. (2025). “Design of an information sharing system for flood response management.” Journal of Internet Computing and Services, Vol. 26, No. 2, pp. 97-108.
Ko, M.-S. (2020). “A study on Korea’s disaster and safety management system from the disaster-site perspective.” Journal of Korean Police Studies, Vol. 22, No. 4, pp. 106-132. 10.24055/kaps.22.4.5
Lee, C.Y., Kim, T.H. (2021). “A study on the standard structure for social disaster and safety accident data analysis.” Journal of the Korea Society of Disaster Information, Vol. 17, No. 4, pp. 817-828.
Lee, Y. (2022). Establishment of KOELECTRA-based Safety Accident Network System and YOLOv5-based Safety Accident Detection for Safety Management. Master Thesis, Sungkyunkwan University.
Lee, Y.K., Kim, W.T. (2023). “A legal review of the introduction process of the National Core Infrastructure system in Disaster and Safety Law.” Crisisonomy, Vol. 19, No. 12, pp. 1-12. 10.14251/crisisonomy.2023.19.12.1
Ouyang, L., Wu, J., Jiang, X., Almeida, D., Wainwright, C.L., Mishkin, P., Zhang, C., Agarwal, S., Slama, K., Ray, A., Schulman, J., Hilton, J., Kelton, F., Miller, L., Simens, M., Askell, A., Welinder, P., Christiano, P., Leike, J., Lowe, R. (2022). “Training language models to follow instructions with human feedback.” Advances in Neural Information Processing Systems, Vol. 35, pp. 27730-27744.
Park, Y.-R., Park, G.-R., Lee, Y.-B., Son, K.-S. (2024). “Generation and utilization of norm-based case data for an AI legal Q&A system.” Journal of Intelligence and Information Systems, Vol. 30, No. 2, pp. 255-273. 10.13088/jiis.2024.30.2.255
Wei, J., Wang, X., Schuurmans, D., Bosma, M., Ichter, B., Xia, F., Chi, E., Le, Q., Zhou, D. (2022). “Chain-of-thought prompting elicits reasoning in large language models.” Advances in Neural Information Processing Systems, Vol. 35, pp. 24824-24837.

Information

Publisher :The Korean Society of Disaster Information
Publisher(Ko) :한국재난정보학회
Journal Title :Journal of the Society of Disaster Information
Journal Title(Ko) :한국재난정보학회논문집
Volume : 21
No :4
Pages :1027-1044
DOI :https://doi.org/10.15683/kosdi.2025.12.31.1027

[1] Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J.D., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., Agarwal, S., Herbert-Voss, A., Krueger, G., Henighan, T., Child, R., Ramesh, A., Ziegler, D.M., Wu, J., Winter, C., Hesse, C., Chen, M., Sigler, E., Litwin, M., Gray, S., Chess, B., Clark, J., Berner, C., McCandlish, S., Radford, A., Sutskever, I., Amodei, D. (2020). “Language models are few-shot learners.” Advances in Neural Information Processing Systems, Vol. 33, pp. 1877-1901.

[2] Choi, H.-C. (2010). “A critical review of the conventional four-phase disaster-management model: Focusing on the response and recovery stages.” Korean Journal of Crisis and Emergency Management, Vol. 6, No. 1, pp. 201-218.

[3] Choi, W. (2020). “Building an intelligent disaster-safety management system based on artificial intelligence.” Journal of the Korean Society of Hazard Mitigation, Vol. 20, No. 1, pp. 127-135. 10.9798/KOSHAM.2020.20.1.127

[4] Chung, K.-S., Jung, W.-S. (2023). “A study on the dataset structure of digital twin for disaster and safety management.” Journal of the Institute of Internet, Broadcasting and Communication, Vol. 23, No. 5, pp. 89-95. 10.7236/JIIBC.2023.23.5.89

[5] Jung, G.S., Jung, W.S. (2022). “Design of social disaster safety platform based on structured/unstructured data.” Journal of the Korea Society of Disaster Information, Vol. 18, No. 4, pp. 715-724.

[6] Kim, J. D. (2010). “An analysis on the order of priority in disaster management policy.” Journal of the Korean Society of Hazard Mitigation, Vol. 10, No. 2, pp. 61-68.

[7] Kim, J., Kim, D.J., Sohn, H.G., Choi, J., Im, J. (2022). “Remote sensing and GIS-based disaster prediction, monitoring, and response (Editorial).” Korean Journal of Remote Sensing, Vol. 38, No. 5-2, pp. 661-667.

[8] Kim, M., Kim, G.Y., Jang, D., Yang, M.S. (2025). “Design of an information sharing system for flood response management.” Journal of Internet Computing and Services, Vol. 26, No. 2, pp. 97-108.

[9] Ko, M.-S. (2020). “A study on Korea’s disaster and safety management system from the disaster-site perspective.” Journal of Korean Police Studies, Vol. 22, No. 4, pp. 106-132. 10.24055/kaps.22.4.5

[10] Lee, C.Y., Kim, T.H. (2021). “A study on the standard structure for social disaster and safety accident data analysis.” Journal of the Korea Society of Disaster Information, Vol. 17, No. 4, pp. 817-828.

[11] Lee, Y. (2022). Establishment of KOELECTRA-based Safety Accident Network System and YOLOv5-based Safety Accident Detection for Safety Management. Master Thesis, Sungkyunkwan University.

[12] Lee, Y.K., Kim, W.T. (2023). “A legal review of the introduction process of the National Core Infrastructure system in Disaster and Safety Law.” Crisisonomy, Vol. 19, No. 12, pp. 1-12. 10.14251/crisisonomy.2023.19.12.1

[13] Ouyang, L., Wu, J., Jiang, X., Almeida, D., Wainwright, C.L., Mishkin, P., Zhang, C., Agarwal, S., Slama, K., Ray, A., Schulman, J., Hilton, J., Kelton, F., Miller, L., Simens, M., Askell, A., Welinder, P., Christiano, P., Leike, J., Lowe, R. (2022). “Training language models to follow instructions with human feedback.” Advances in Neural Information Processing Systems, Vol. 35, pp. 27730-27744.

[14] Park, Y.-R., Park, G.-R., Lee, Y.-B., Son, K.-S. (2024). “Generation and utilization of norm-based case data for an AI legal Q&A system.” Journal of Intelligence and Information Systems, Vol. 30, No. 2, pp. 255-273. 10.13088/jiis.2024.30.2.255

[15] Wei, J., Wang, X., Schuurmans, D., Bosma, M., Ichter, B., Xia, F., Chi, E., Le, Q., Zhou, D. (2022). “Chain-of-thought prompting elicits reasoning in large language models.” Advances in Neural Information Processing Systems, Vol. 35, pp. 24824-24837.

Journal of the Society of Disaster Information ISSN:1976-2208(Print) 2671-5287(Online) 한국재난정보학회논문집

All Issue