Acmqueue November December 2019 issue is about human factors in software - John Allspaw curated

An incident is a way the system tells you that something isn’t working the way you think it does. Helps to focus a roadmap :)

Post incident review docs are written to be filed, not read

Learning requires (Conditions where learning can happen)

  • psycological safety
  • blamelessness

Hindsight: simplify events to a comfortable narrative. How we deal with an unpleasant surprise. This loses information. Multiple perspectives are good if you can retain them.

Before an incident, conditions are setup for an operational surprise to happen

Who tells the story decides

  • what is remembered
  • what is fixed
  • how the story is told

How do you assemble all the different perspectives of people that participated in an an event into something greater than the individual pieces?

Techniques in the new world

  • 1:1 interviews
  • “Synthesis”
  • What was difficult, surprising
  • How did it get fixed
  • What mysteries still remain

Goal is to fix people’s mental models. Bit of prep for the sharing meeting. Wait. Let material soak. Then decide what to do …

Record who was in the room during the post incident review

How do you write a document to be read? Evoke an emotional response. A feeling in your gut that other people have felt