No theory forbids me to say "Ah!" or "Ugh!", but it forbids me the bogus theorization of my "Ah!" and "Ugh!" - the value judgments. - Theodor Julius Geiger (1960)

Safety as a network

Le Coze, J.C., Dupré, M. (2022), Safety as a network, digital and global reality, in: Safety Science, Vol. 156, 105896.  

This paper by Le Coze and co-author, sociologist of work, Dupré, argues that today’s new age of work (globalisation, digitalisation/automation, financialisation, standardisation, externalisation, and more self-regulatory oversight) redefines safety as a multilevel, networked phenomenon. Using an ethnographic case from a French chemical plant, it shows how macro trends reconfigure technology, organization, and daily work. This changes how safety is produced on the ground.

Method

Ethnography (2011–2012): several weeks on site across seven visits; ~40 interviews; observation of shifts, control room, maintenance, loading/unloading; document analysis; and analysis of recent incident investigations.

Site: ~100 employees (+ contractors), part of a 3,500-person multinational.

What changed between the 2000s until the early 2010s

Firstly, digitalisation: “The advent of informational infrastructures”

- Manual reactors became highly automated; team size shrank (≈12 → 7).

- Work changed from physical tasks to cognitive supervision + lab testing.

- A new central actor emerged: the IT engineer who programs control-system recipes (Foxboro). Operators’ autonomy narrowed to parameters the software allows.

- Interface gaps (alarm floods, outdated screen schematics, hidden sensors) made field checks indispensable—operators reconcile what screens show with what the plant does.

Secondly, standardisation, self-regulation, and financialisation: “Into the matrix”

- The firm moved to a matrix, region-level structure (4 regions; functional VPs). - Local plant autonomy decreased; corporate visibility and comparability increased.

- HSE was formalised and scaled: corporate “minimum requirements,” LOPA/PHAs, Management of Change (MOC), incident reporting; alongside tighter national regulation after major accidents.

- Result: more bureaucracy and reporting, new English-speaking mid-managers to interface with corporate, and occasional misalignment between corporate priorities and site realities—yet overall a stronger process-safety baseline.

Thirdly, externalisation: “WaterT & Co.”

- Broader reliance on contractors (water treatment, boilers/steam, gas, inspections, logistics, risk analysis).

- Contracts add legal/admin work; HSE integration of contractors becomes a major task.

- The outsourced water-treatment station illustrates dependency risks: mixed vendor performance affected production pacing and safety coordination, prompting tighter daily integration with plant operations.

Consequences for safety-as-practice

Where tasks and identities are concerned, operators become information workers; less physical solidarity, more individual screen work; IT expertise gains power.

Where power and coordination are concerned: decision-making redistributes upward (matrix) and outward (vendors); plant managers’ autonomy narrows; bureaucratic load rises.

Where interactions are concerned: more cross-functional, cross-site, and inter-organizational coordination is needed to keep the system safe.

Takeaways

Safety emerges from a network of digital systems, corporate standards, regulators, and contractors, not just from the plant’s internal practices.

Standardisation and digitalisation improve baseline control but introduce new failure modes (alarm noise, stale interfaces, recipe backlog) and coordination burdens.

Externalisation can bring needed expertise yet creates dependencies that must be actively managed (contracts, competence, day-to-day integration).

Implications & outlook

Expect more data-driven, remote-supervised operations, greater reliance on IT (and cybersecurity as a safety concern), and potential de-globalisation changes in value chains, each re-wiring the network of safety.

Ethnography remains crucial to see how these macro trends actually land in daily work and where compensatory professional expertise still carries the system.

Limits

Single-site, time-bound ethnography; findings are analytically generalisable (via the integrated framework) rather than statistically so.