A RAG-Based Decision Support System for Occupational Exposure Assessment of VOCs: Design, Implementation, and Evaluation

Authors

  • Hadis Vahedi Department of Occupational Health and Safety Engineering, School of Public Health and Safety, Shahid Beheshti University of Medical Sciences, Tehran, Iran
  • Mehrnoush Shamsfard Natural Language Processing (NLP) Lab, Faculty of Computer Science and Engineering, Shahid Beheshti University, Tehran, Iran
  • Ehsan Toofaninejad E-Learning in Medical Sciences Department, School of Medical Education and Learning Technologies, Shahid Beheshti University of Medical Sciences, Tehran, Iran
  • Zahra Vatankhah Mohammadabadi Natural Language Processing (NLP) Lab, Faculty of Computer Science and Engineering, Shahid Beheshti University, Tehran, Iran
  • Fatemeh Dorfeshan Student Research Committee, School of Public Health and Safety, Shahid Beheshti University of Medical Sciences, Tehran, Iran.
  • Somayeh Farhang Dehghan Environmental and Occupational Hazards Control Research Center, Research Institute for Health Sciences and Environment, Shahid Beheshti University of Medical Sciences, Tehran, Iran.

DOI:

https://doi.org/10.18502/jhsw.v16i2.22915

Keywords:

Occupational health, Chatbot, Retrieval-Augmented Generation (RAG), NIOSH standards, Volatile organic compounds (VOCs), Decision support system (DSS)

Abstract

Introduction: Assessing exposure to volatile organic compounds (VOCs) in workplaces requires timely access to accurate, standardized guidance on sampling and analysis. This study aimed to develop and evaluate a decision-support chatbot, based on the Retrieval-Augmented Generation (RAG) architecture, to assist occupational health professionals in selecting appropriate sampling and analysis methods for VOCs in accordance with NIOSH guidelines.

Material and Methods: This research was of an applied-developmental type. In this study, relevant NIOSH methods for sampling and analysis of VOCs were extracted, structured, and organized into a knowledge base. User needs were identified through semi-structured interviews with occupational health professionals. Based on this data, a generative AI-based chatbot was designed and implemented using the RAG architecture. The system performance was evaluated using expert evaluation and a chatbot usability questionnaire.

Results: The developed chatbot received an overall user-friendliness score of 3.88 out of 5, indicating good user acceptance. Participants noted improved access speed for standard technical information and the high transparency of the system’s responses. Expert evaluation also confirmed the scientific validity of the retrieved content and its compliance with NIOSH standards.

Conclusion: The proposed RAG-based decision-making chatbot provides an effective and innovative tool to support occupational exposure assessment to VOCs. Integrating generative AI with standard occupational health guidelines can improve the efficiency of decision-making and access to expert information in practice.

Published

2026-10-07

Issue

Volume (16) issue (2)

Section

Articles