A Comparative Study on the Consistency of Construction Site Safety Improvement Measures between Generative AI and Human Safety Managers

doi:10.15683/kosdi.2026.6.30.644

All Issue

2026 Vol.22, Issue 2 Preview Page Next Page

Original Article

A Comparative Study on the Consistency of Construction Site Safety Improvement Measures between Generative AI and Human Safety Managers 생성형 AI와 안전관리자의 건설현장 안전개선조치 일치도 비교 연구: 윤태경¹ㆍ윤형구²*
Tae-Kyeong Yoon¹, Hyung-Koo Yoon²*; ¹Head of Safety and Health, Shindonga Construction Co., Ltd., Seoul, Republic of Korea
²Professor, Department of Disaster and Safety Engineering, Daejeon University, Daejeon, Republic of Korea

30 June 2026. pp. 644-653

PDF

Abstract

Purpose: This study aims to compare and analyze the level of agreement between safety improvement measures proposed by generative AI (ChatGPT) and those implemented by safety managers in actual construction sites, and to verify whether they meet legal standards and practical requirements. Method: Cases of safety issues identified at construction sites were collected, and improvement measures suggested by AI were compared with those implemented by safety managers under the same conditions. The analysis was conducted based on detailed agreement rates according to risk types and issue categories. Both quantitative and qualitative comparisons were performed to evaluate the effectiveness of AI-generated suggestions. Result: The results showed the highest level of agreement in fall-related hazards and facility-related safety issues. In contrast, relatively lower agreement rates were observed in pinching hazards and management system improvement cases. Conclusion: The findings suggest that generative AI can be utilized as a supportive tool for construction safety management. Furthermore, the results of this study may serve as foundational data for the future development of AI-based safety management support systems.

Keywords

Generative AI

ChatGPT

Construction Safety

Improvement Measures

Consistency

연구목적: 생성형 AI(ChatGPT)가 제안한 건설현장 안전 개선조치와 안전관리자가 실제현장에서 수행한 개선조치 간의 일치율을 비교·분석하여 법적 기준과 실무적 요구사항에 충족하는지에 대한 검증을 목적으로 한다. 연구방법: 건설현장의 지적사항 사례를 수집하고, 동일한 조건에서 AI 개선조치 제안과 안전관리자 조치를 비교하였다. 분석 기준은 위험·지적유형별 세부 일치율이며, 정량·정성적 비교를 통해 AI 제안의 실효성을 검토하였다. 연구결과: 추락과 시설에서 가장 높은 일치율을 보였으며, 끼임·관리체계 개선 분야 지적에서는 상대적으로 낮은 일치율이 나타났다. 결론: 생성형 AI의 건설안전 관리 보조 도구로서의 활용 가능성과 향후 AI 기반 안전관리 지원 시스템 개발의 기초자료로 활용될 수 있다.

키워드

생성형 AI

ChatGPT

건설안전

개선조치

일치율 분석

References

Gerard, S. (1965). The SMART Retrieval System-Experiments in Automatic Document Processing. IOS Press, Amsterdam.
Kim, T.-Y. (2024). Announcement of the Results of the 2023 Industrial Accident Supplementary Statistics: Status of Fatal Accidents Subject to Accident Investigation. Press Release, Ministry of Employment and Labor, Sejong.
McKinney, W. (2011). “Pandas: a foundational Python library for data analysis and statistics.” Python for High Performance and Scientific Computing, Vol. 14, No. 9, pp. 1-9.
Ministry of Government Legislation (2021). Act on the Punishment of Serious Accidents (Act No. 17907, Jan. 26, 2021). National Law Information Center, Sejong.
Oh, E.-K. (2025). Announcement of the Status of Fatal Occupational Accidents Based on Bereaved Family Benefit Approval in 2024. Press Release, Ministry of Employment and Labor, Sejong.
OpenAI (2023). GPT-4 Technical Report. OpenAI, Technical Report, San Francisco, CA, USA.
Park, E.L., Cho, S. (2014). “KoNLPy: Korean natural language processing in Python.” 26th Annual Conference on Human & Cognitive Language Technology, Daejeon, Vol. 6, pp. 133-136.
Salton, G. (1958). Automatic Information Organization and Retrieval. Ph.D. Dissertation, Harvard University, Cambridge, USA
Son, P.-H. (2022). Announcement of the Status of Fatal Occupational Accidents in 2021. Press Release, Ministry of Employment and Labor, Sejong.

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 : 22
No :2
Pages :644-653
DOI :https://doi.org/10.15683/kosdi.2026.6.30.644

[1] Gerard, S. (1965). The SMART Retrieval System-Experiments in Automatic Document Processing. IOS Press, Amsterdam.

[2] Kim, T.-Y. (2024). Announcement of the Results of the 2023 Industrial Accident Supplementary Statistics: Status of Fatal Accidents Subject to Accident Investigation. Press Release, Ministry of Employment and Labor, Sejong.

[3] McKinney, W. (2011). “Pandas: a foundational Python library for data analysis and statistics.” Python for High Performance and Scientific Computing, Vol. 14, No. 9, pp. 1-9.

[4] Ministry of Government Legislation (2021). Act on the Punishment of Serious Accidents (Act No. 17907, Jan. 26, 2021). National Law Information Center, Sejong.

[5] Oh, E.-K. (2025). Announcement of the Status of Fatal Occupational Accidents Based on Bereaved Family Benefit Approval in 2024. Press Release, Ministry of Employment and Labor, Sejong.

[6] OpenAI (2023). GPT-4 Technical Report. OpenAI, Technical Report, San Francisco, CA, USA.

[7] Park, E.L., Cho, S. (2014). “KoNLPy: Korean natural language processing in Python.” 26th Annual Conference on Human & Cognitive Language Technology, Daejeon, Vol. 6, pp. 133-136.

[8] Salton, G. (1958). Automatic Information Organization and Retrieval. Ph.D. Dissertation, Harvard University, Cambridge, USA

[9] Son, P.-H. (2022). Announcement of the Status of Fatal Occupational Accidents in 2021. Press Release, Ministry of Employment and Labor, Sejong.

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

All Issue