The case against safety conversations
"Why don't you ask questions directly about safety? All your questions are about work, about what's difficult. None of them are about hazards, controls, dangers."
The training participant seemed equally frustrated and intrigued. And I think I get why. Asking about how work is going seems unstructured, and perhaps even evasive. If you're there to talk about safety, why not just talk about safety?
My honest answer is: because the safety door opens into a room most people have already tidied.
Everyone who has ever been asked about safety at work knows what a good answer looks like. Hazards identified, controls in place, near misses reported, procedures followed. The conversation has a shape before it starts, and that shape produces a particular kind of response, which tends to be accurate, compliant, and largely uninformative about what's actually happening.
The ‘work door’ opens somewhere different.
Work is full of sensitivities, fragilities, and challenges that don't have a category in any safety system. The task that takes longer than the schedule allows. The tool that isn't quite right for the job but close enough. Or the two competing priorities that can't both be met, so something gives quietly, locally, without anyone formally deciding that it should. None of that gets named in a conversation that announces itself as being about safety. It doesn't fit through that door.
But ask someone what's been difficult lately, what they're paying attention to, where things get a bit unpredictable, and that material starts to surface. Not because people were hiding it, but because nobody had asked for it before.
There's a structural reason this works. Safety talk triggers a particular cognitive mode: assessment against standard. The person being asked is implicitly being compared or evaluated: are things as they should be? That mode produces compliance answers because compliance answers are what the situation calls for.
Work talk triggers something different. It invites description of experience rather than assessment of performance. And description of experience is where the actual information lives; the stuff that is true before it has been filtered through what the system wants to hear.
The very conversations designed to surface risk can prevent it from appearing, because they tell people what kind of answer is expected before they've had a chance to say something surprising.
I have asked about hazards and controls during a safety conversation and watched someone carefully give me the right answer. I've also asked someone what's been tricky lately and ended up, without planning to get there, in a conversation about a fracture in the team that had been building for months, a piece of equipment being nursed past its useful life. Perhaps not labelled as safety things, but it’s not difficult to imagined how they woulod contribute to an accident.
Safety arrived into those conversations. Through the work, rather than announced at the door.
Work is full of sensitivities and fragilities. Follow them with genuine curiosity and you end up somewhere interesting. Often somewhere that matters for safety. Sometimes somewhere that matters for other things entirely: well-being, productivity, capability, organisational health.
I don't decide in advance where the conversation should go. I let work lead me there.
So the case against safety conversations isn't really a case against safety. It's a case against conversations that declare their subject before they've earned the right to go anywhere unexpected.
The best safety conversation is often one that doesn't announce itself as a safety conversation at all. It starts with work, with what's difficult, what's unpredictable, what people are navigating, and arrives at what matters by following what's real.
And that’s choosing the instrument most likely to surface what you actually need to know, rather than the one most likely to confirm what you already believe.
Daniel
PS
Mary Conquest at Safety Labs did a 1h interview with me about my book. Have a look and listen at: https://www.youtube.com/watch?v=zeodX0-GXCg&t=1310s