Learning from Nature: A Biomimetic Network Framework for Occupational Health and Safety Research

Abstract

Occupational health and safety research has long relied on two ways of understanding harm. One traces the chain of conditions behind an accident, and the other forecasts the outcome of an incident from large bodies of recorded cases. Both treat the factors surrounding an incident as inputs pointing toward a result. Neither was built to ask how the categories of risk in a whole body of records are organized in relation to one another, independent of any single outcome. This paper addresses that gap through biomimicry, taking the way living systems organize their interactions as a template for reading how risk is organized in safety data. Ecologists have long described how two distinct sets of entities connect through bipartite networks, and they have developed a compact set of measures for reading the structure of those networks, covering whether interactions concentrate in a shared core or separate into compartments, which categories are specialized and which dominate, and how a system responds when its parts are removed. Occupational safety data fit this biological template, because the categorical fields that safety agencies record pair naturally into the same two-set form. Transferring these ecological measures to safety records, this study sets out how the measures map onto safety questions and illustrates their use across construction, mining, manufacturing, and healthcare. A worked example then applies the full procedure to a public dataset of 158,396 healthcare injury records, building the bipartite matrices and reading their nestedness, modularity, specialization, and robustness against a null model, which shows that the approach runs end to end on real administrative data. The study presents this as a proposed framework and a demonstration of its feasibility rather than a validated method and argues that this nature-derived structural reading complements rather than replaces the causal and predictive traditions.

The authors' abstract, as published at the source. Biomimetics, 2026 · DOI ↗