Most organisations talk about responsibility as if agreement on values like “safety first!” guarantees safer outcomes. The reality is different. What really matters is how we observe ourselves and communicate that observation. In other words, “working safely” happens when communication about work becomes a dynamic, revisable tekst. The organisation is describing itself as it learns.

1️⃣ From danger to risk

We still hear a lot of talk about accidents being “freak accidents” or “bad luck”. In his book “Risk: A Sociological Theory”, Niklas Luhmann separated danger (harm from outside) from risk (harm from our own decisions). This is important, because decisions about modern technology and processes convert external dangers into risks.

Firstly, it’s important to treat every incident, near-miss and waiver primarily as a decision outcome. We can do this by keeping decision logs, pre-mortems and trade-off records so that learning targets organisational choices rather than people at the sharp end.

Secondly, don’t just study what went wrong. Study how things normally go right under variable conditions. Capture the small adaptations, workarounds and quiet recoveries that keep operations stable. Because this is resilience in action: the capacity to succeed under both expected and unexpected conditions!

2️⃣ From values to verification

Ethics committees and pledges often stall at implementation difficulties. So, it’s important to replace vague commitments with self-observing mechanisms. For example:

• Write safety cases as living documents (with owners, review cadence and edit history);
• Run routine double-loop reviews and ask: “Did we choose the right constraints?”
• Publish “Why we accepted this risk” notes alongside approvals.

Then go one step further: add resilience cases. Document how the system recovers instead of only how it prevents. Do this by showing the buffers and fallback plans, and who can re-prioritise when surprises hit. Resilience needs verification too, as it proves that your system can bend without breaking.

3️⃣ From announcements to feedback

Mass communications drift toward drama and amnesia. Accidents are hot for a short time, and then organisations slowly forget.

What we can do instead:

• Build quiet, continuous feedback loops (e.g. blameless near-miss reporting);
• Maintain weak-signal dashboards that capture even tiny deviations;
• Run red-team (or: “learning team”) drills to challenge your own narratives;
• Hold post-incident reviews that focus on system conditions isntead of heroics or blame; and
• Debrief successful recoveries and normal work just as often as you review failures;
• Ask teams, “What helped us get this right?”, and codify those answers so adaptation becomes teachable

Conclusion

Doing all this creates a safety function that doesn’t just promise responsibility, but documents how the organisation learns, decides and adapts. That self-description, “resilience made visible”, is your most powerful control.

To get started: ask next week:

• Where is our risk actually a product of our decisions?
• Which texts (procedures, cases and reviews) prove we are observing ourselves?
• What did we change after the last near-miss, and where is that written down?
• And: what went right under pressure, and how are we preserving that capacity?

Safety improves when communication becomes traceable learning; when both success and failure are recorded as evidence of adaptation. So, let’s make our systems transparent enough to teach us how they stay resilient!