The Field Guide to Understanding Human Error

Read this a while ago - a few things I think now that I attribute to the book:

systems work most of the time, and it's because people generally can figure out how to do the right thing
sometimes systems fail, because people did not figure out how to do the right thing
* when a tense situation is unfolding, the people involved are making choices that make sense to them in that context. If the choice turns out to be wrong, that is regrettable, but also, "it was the right idea at the time."

A systemic view on human errors. The title is actually a bit misleading, since the book states that “human error” is caused by an interplay of incentives, culture, authority/responsibility and co-existing pressures, and rarely by single people who behave wrong.

Summary:

Dekkers argument is that "should not have done" makes sense only in hindsight, with the information an outsider has after the error has happened. Most people want to do their job right, and thus they usually do what makes sense in their situation, weighting to options and the different pressures (economic, social, safety…).

Dekker argues, that "Human Error","Action X was wrong", "We should just have…" covers up the complex ways in which problematic situations build up and particularly, how the larger organizational system is tied to it: Errors can tell you more about the organization and its problems than about the "wrong" individual: A person used a wrong tool and as a result the machine broke? "Human Error" and shoddy work! Or rather it is because of this: The workers can go home when work is done, their tools are stored far away, so they clean up before the shift ends so they can leave. If now, unexpected, more work comes in, they will use whatever tools are there, otherwise everyone suffers a loss of wage. The latter is more interesting, more understandable than "human error". It is also less comfortable for the organization, which now plays a part in the problems rather than that one badly behaving person.

This is something that can't be quick-fixed by more procedures (how to interpret them?), technology (how does it work? When?) and some trainings (since the people close to the error might not be the source of it).

Instead, Dekker suggests a slow approach, starting with a detailed, observable-facts based account of the error with the interpretations of those who were close to it and why their actions made sense to them at that time, followed by a review of how the organization can learn from it.