A systematic approach for identifying drivers of critical safety and establishing their hierarchy
Bibliographic record
Abstract
Abstract Learning from incidents is a crucial step in preventing and mitigating adverse events. Incident databases offer valuable insights for safety management improvements by cause and contributing factors. However, extracting meaningful information from incident investigation reports poses a significant challenge. This study introduces a data‐driven methodology to assess drivers of critical safety (DCS), which is essential for enhancing the safety of the process industries and protecting workers and the environment. Natural language processing (NLP) can offer automated, actionable insights from incident investigation reports. This automation is important in identifying DCS from incident reports to ensure proactive prevention and effective mitigation of risks, thereby protecting assets, workers, and the environment. Based on lagging safety indicators (causes or contributing factors), we aim to develop leading safety improvements to enhance the safety management system. A crucial step involves developing a DCS hierarchy to assess their role within the overall safety management framework. This hierarchy quantifies the driving and dependence power of each driver. The former refers to the number of drivers affected by each driver, while the latter determines the number of drivers impacted by each driver. This hierarchy facilitates resource allocation and determines each driver's effectiveness in safety management. The tool is developed and trained using the publicly available CSB database, a comprehensive source of incident investigation data. To further verify the model's effectiveness, it is tested and verified on an unseen database of 26 release incidents released by CSB in January 2025. The model successfully identifies the DCS responsible for each incident.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.011 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".