In the field of industrial safety and risk management, diagnosing human factors in operational events remains one of the greatest challenges. Traditionally, this process relied on subjective expert interpretation of narrative reports and formal guidelines, but with high time costs and considerable variability in outcomes. The arrival of data-driven approaches and large language models (LLMs) promised more efficient automation, but in practice these systems lacked structured reasoning, aligned weakly with diagnostic guidelines, and generated logically inconsistent conclusions. This is where G-SHARE comes in—a structured reasoning framework based on guidelines that transforms diagnostic directives into an auditable and reproducible workflow.
G-SHARE is not just another tool; it represents a paradigm shift in how organizations approach human event diagnosis. Its architecture is divided into three key phases: evidence extraction from the report, step-by-step diagnostic reasoning following a nine-step guideline, and a post-hoc consistency repair mechanism that logically validates the conclusions. This approach ensures that each step of the reasoning is documented and can be audited, which is essential in critical industries where misinterpretation can have serious consequences. By operationalizing a formal guideline—in the case study, the CNNP nine-step guideline—G-SHARE guarantees that the diagnosis is not only accurate but also transparent and justifiable.
Test results on a real dataset of human event reports from the Chinese nuclear industry show that G-SHARE significantly outperforms one-shot prompting and traditional machine learning baselines. The most powerful version of the framework achieved the best overall accuracy and a higher macro-F1, confirming that structured reasoning and consistency validation are critical factors for robust diagnosis, especially under weak prompting conditions. These findings not only validate G-SHARE's utility but also open the door to its application in other sectors where diagnostic reliability is paramount.
From a business and technology perspective, G-SHARE perfectly illustrates how transforming expert guidelines into computational workflows can revolutionize critical processes. At Q2BSTUDIO, we understand that the key is not just implementing artificial intelligence, but doing so in a structured, auditable, and business-aligned manner. That's why we offer AI services that enable companies to design similar reasoning systems adapted to their own domains. Furthermore, integrating these frameworks with cloud platforms like AWS or Azure—through our cloud AWS/Azure service—provides the scalability and flexibility needed to handle large data volumes and run complex models in real time.
But the value of G-SHARE goes beyond pure diagnosis. Its modular, rule-based architecture can be applied to other areas such as audit process automation, anomaly detection in industrial systems, or regulatory compliance validation. At Q2BSTUDIO, we combine this vision with our expertise in custom software, creating solutions that not only analyze data but also generate clear, actionable explanations for decision-makers. Cybersecurity also plays a crucial role: any diagnostic system handling sensitive information must be protected. Our cybersecurity solutions ensure that data and models are safe from unauthorized access, maintaining process integrity.
Another relevant aspect is G-SHARE's ability to integrate with Business Intelligence tools like Power BI, enabling intuitive visualization of diagnostic results. At Q2BSTUDIO, we offer BI/Power BI services that turn data generated by these frameworks into dynamic dashboards, facilitating trend monitoring and identification of recurring patterns in human events. The combination of structured reasoning and data visualization is a powerful tool for continuous improvement in any organization.
Beyond the nuclear industry, sectors such as aviation, healthcare, energy, or finance can benefit from an approach like G-SHARE. The key lies in the ability to transform expert guidelines—often implicit or scattered across documents—into computational workflows that integrate artificial intelligence, natural language processing, and logical validation. At Q2BSTUDIO, we work with our clients to develop custom AI agents that not only execute tasks but reason about them, explaining each step of the process.
Ultimately, G-SHARE demonstrates that the future of human factor diagnosis lies in structuring, auditability, and the integration of advanced technologies. It is not about replacing experts, but empowering them with tools that amplify their analytical capacity and reduce errors. At Q2BSTUDIO, we are committed to that vision, offering a complete ecosystem of services—from custom software to cloud, cybersecurity, BI, and AI—that enables companies to make the leap toward intelligent, reliable, and scalable diagnosis.
