Three books that rewired how I think about risk

People keep recommending me risk-management books full of Greek letters. I've read some. They're useful in the way a dictionary is useful. The three books that actually changed how I sit down in front of a model have no Greek letters in them at all — they're about mountains, mushrooms, and a surgeon with a clipboard.

Touching the Void — Joe Simpson

A climbing-accident memoir, and the clearest account of path dependence I have ever read. Simpson survives a broken leg, a crevasse and a cut rope not by making one brilliant decision but by consistently refusing to make the one irreversible one. He crawls toward a series of small, near-term, achievable targets — that rock, then the next rock — for three days.

You gotta keep making decisions, even if they're wrong decisions. If you don't make decisions, you're stuffed.

The translation into anything I do is embarrassingly direct. A position, a model, a career step: the question is rarely is this right. It's if this is wrong, can I still move? I think about that line every time I catch myself frozen over a spreadsheet that already answered the question I'm still asking.

Entangled Life — Merlin Sheldrake

Nominally a book about fungi. Actually a book about systems that survive by being decentralised, which is most of the interesting systems. Mycelial networks have no centre, no plan and no manager, and they route resources around damage with a competence that makes a supply chain diagram look naive.

I read this the same semester I was studying network resilience for my operations major, and the two rearranged each other. The formal version says redundancy costs money and buys robustness. Sheldrake's version says the redundancy is the organism. A network with no slack in it isn't efficient, it's one bad node from being nothing.

It also permanently ruined the word “optimise” for me. Optimised for what, over what horizon, surviving what?

The Checklist Manifesto — Atul Gawande

Gawande's distinction is the one I've quoted most often: there are failures because we don't know enough, and failures because we didn't properly use what we already knew. The first kind is a research problem. The second kind is far more common, far more embarrassing, and — this is his whole point — fixable with something as unglamorous as a list.

Every mistake I made in my first modelling internship was the second kind. Not one of them was a gap in knowledge. They were a forgotten sign convention, a range that didn't extend to the new row, a stale assumption I'd have caught if anyone, including me, had read the thing aloud.

I now run a short one before anything leaves my hands. It has fewer than ten items and it has caught something roughly every third time, which is a return no model I've built comes close to.

What they have in common

None of these books uses the word “risk” the way a finance curriculum does. Between them they gave me three questions I now ask of almost anything:

• Is this reversible? — Simpson. Irreversibility is the only thing that genuinely can't be recovered from.

• Where is the slack? — Sheldrake. A system with none is not efficient, it's brittle and hasn't been tested yet.

• Is this ignorance or ineptitude? — Gawande. They have completely different fixes, and I default to assuming the first when it's almost always the second.

The Greek-letter books tell you how to measure risk once you've agreed what it is. These three are about the part before that, which in my limited experience is where things actually go wrong. Next on the pile is something about air-accident investigation, on the theory that any field that publishes its failures in public has learned something the rest of us haven't.