(101k words) Somewhat disappointing. Dekker focuses on how the occasional rare disaster is very complex, which they can be, but never gets off his high horse, instead spending a whole book talking about how everything is terribly terribly complicated.

He sets up a strawman in which attempts to improve things evidences a naive and naturally wrong "Newtonian worldview" in which tragic cases of reductionism run amok guarantee disaster, and particularly mocks the swiss-cheese model of failure. The problem is, the more he rails against it and runs through the stereotypical case-studies like Challenger, the more apt that model seems, and the less Dekker seems to be trying to understand 'Newtonian' views or acknowledge that improvements are possible, that drift into failure is not inevitable. For example, in discussing the recommendations produced by an aircraft investigation, like making a hatch easier for mechanics to see into, he asks rhetorically how all these fatally flawed designs and procedures could possibly have gone unnoticed before the accident; the answer is, of course, that airplane accidents are vanishingly rare and so a contributing cause can be responsible for a large fraction of the current accident rate and that flaw be unnoticed before the accidents - if only 0.001% of planes crash and a too-narrow hatch is responsible for a full 10% of all crashes, then the hatch is a risk of only 0.0001%, and it's entirely understandable that no one would notice that, and also true that fixing the hatch is worthwhile!

And one might think that he would devote great attention to the steady decline in aircraft fatalities; I kept waiting for that to come up, and on pg150, it finally does in the mouth of one of his students, and his response is... global warming. I'm not making this up, that is what he says, he ignores the fatality question and starts talking about global warming. Give me a break. More generally, there never seems to be an economic perspective taken: what is the right number of disasters? Why must there be zero disasters and any disasters are a failure of 'Newtonian' thinking?

I experienced a similar bit of shock to see how much space he devoted to the Gaussian copula; the Gaussian copula, for those who don't recognize it, is an obscure bit of financial modeling which, after the housing bubble burst in 2007/2008, a few people blamed its assumption of independence for causing underestimates of the probability of housing prices falling, but you haven't heard of it since because it quickly became apparent that this was a ridiculous theory which didn't explain anything and was picking out an incidental surface feature while ignoring the real underlying drivers of the crisis.

Overall, I don't think I learned much from this. If you're interested in disasters and complex systems, you're probably better off reading more standard texts like Feynman's "Appendix F - Personal observations on the reliability of the Shuttle" or even just random post-mortems like the 2001 report "The explosion of No. 5 Blast Furnace, Corus UK Ltd, Port Talbot".